Hacker News, Distilled

Cultural loss and value of MTV News

• Many see the takedown as a serious loss of music and youth‑culture history (interviews, breaking news like major artist deaths, series like “True Life”).
• Others downplay MTV News’ importance, arguing it wasn’t widely used or high‑quality journalism, highlighting how subjective “what’s worth saving” is.
• Some note that even “trivial” or lowbrow content can be culturally important in hindsight.

Who should archive: Internet Archive vs. government

• One view: Internet Archive (IA) should get major public funding; storage is cheap and cultural loss is huge.
• Counterview: archiving implicates rights of subjects, authors, and future uses (e.g., LLM training), so it should be overseen by public institutions with clear accountability.
• Compromise positions:
  • IA should remain independent to avoid political interference; governments could run parallel mirrors.
  • IA is fragile and not “too big to fail”; people fear lawsuits, buyouts, or policy pressure.
• Examples raised: Library of Congress’ very selective web archiving; legal‑deposit systems in the UK/Canada; perma.cc as an institutional solution.

Copyright, fair use, and archival rights

• Strong tension between “copyright is too long and blocks preservation” vs. “copyright needs strengthening; archiving can be handled via existing exceptions.”
• Some propose:
  • Shorter terms or differentiated terms by content/value.
  • Copyright tied to stewardship: lose exclusive rights if you don’t preserve and provide access.
  • Clearer legal exceptions when originals vanish or rights‑holders take them offline.
• Journalists’ archival rights are flagged as a labour issue; losing employer sites erases portfolios.

Piracy and informal preservation

• Several argue piracy has been crucial to preserving unaltered works (e.g., theatrical film cuts, TV with original music).
• Examples include fan restorations, private TV recording projects, and personal VHS archives.
• View that “piracy is core to historical preservation” is contested but strongly voiced.

Link rot, paywalls, and personal archiving

• Growing frustration with:
  • Content disappearing or going behind new login/paywalls after decades.
  • “Link rot” making old citations and bookmarks useless.
• Suggested mitigations:
  • Routine use of IA or similar at bookmark time.
  • Browser extensions (e.g., full‑page savers), personal HTML archives, private git repositories.
• Some predict a future “information black hole” for this era despite the digital boom.

Scale, cost, and technical issues

• For text and images, many argue storage is cheap enough to “keep almost everything”; selection introduces bias and risk.
• Video is harder: archiving all platforms (e.g., YouTube, TikTok) is massive but maybe feasible if highly selective.
• Debate over whether shortened copyright terms materially change feasibility, given need for durable, replicable archives over decades.

Censorship, editing, and control of archives

• Concern that centralized or corporate archives can be altered or sanitized (e.g., editing old cartoons, removing smoking, modifying soundtracks).
• Fears extend to:
  • Political interference in state archives.
  • Rights‑holders using takedowns to reshape history.
  • Even IA being pressured to remove specific items.
• Some call for decentralized or blockchain‑based archival; others question whether blockchain adds much beyond traditional hashing and funding models.

Work style and the original anecdote

• Commenters note that the engineer in the story partly worked from home to avoid interruptions, seeing it as an early “healthy split” remote/on-site model.
• The narrative is praised as charming and motivating: a boss demanding “better” and a developer rising to the challenge instead of sulking.

How rounded rectangles were implemented

• Several ask for more technical depth than the original story provides.
• Linked resources suggest QuickDraw used an algorithm akin to the midpoint circle algorithm, drawing arc octants and stitching them to straight edges, likely integrated with existing region/fill logic.
• Others speculate that for common cases, precomputed corner masks may have been used for speed.

Design, aesthetics, and usability

• Many praise rounded corners as making interfaces look modern, cheerful, and more approachable.
• Some argue they are more than fashion: rounded borders help visually group elements and convey hierarchy (gestalt, accessibility).
• Others see corner styles as cyclical fads, pointing to repeated shifts between rounded and sharp UI in operating systems and browsers.

Squircles and geometric nuance

• Discussion covers squircles and superellipses in modern UI, including quintic superellipses with smooth tangency (C2 continuity).
• Some argue these curves matter for 3D product highlights but are visually subtle for icons.
• There is debate over whether Apple-style rectangular elements are true superellipses or mainly standard rounded rectangles.

Hardware vs software corner rounding

• Some claim laptop and phone screen corner rounding is purely software (black pixels).
• Others counter with examples of panels physically manufactured with rounded masks or cutouts, implying no pixels exist there.

Broader reflections: leadership, culture, and history

• Commenters revisit the co-founder’s design taste and “liberal arts + technology” mantra, contrasting genuine product vision with later imitators.
• There is debate over whether their behavior was visionary leadership or narcissistic toxicity, with acknowledgment of harsh treatment of employees.
• Nostalgia threads compare older OS UIs (with clear 3D affordances, tighter spacing) to today’s flatter, more spacious, often less responsive designs.

Who’s Buying Distressed CRE Loans and Why

• Buyers include specialized investors, other real estate firms, and entities willing to act as landlords or work out distressed debt.
• Motivations: buy loans at a discount, bet on sector recovery, restructure at higher rates, or foreclose to obtain buildings cheaply.
• For some, taking possession of the building is “plan A”, not a worst case.
• Banks are constrained by capital and regulatory requirements, so even if loans have long‑term value, offloading them improves their balance sheets.

How Bad Can the Loans Get?

• One side: secured CRE loans in the US are rarely “worth nothing” because the property backs them, and even defaults can be worked out via “special servicing.”
• Other side: second/subordinated liens can indeed go to zero when values fall and legal/administrative costs exceed recoveries; this is more common late in cycles and in non‑recourse states.
• Debate over borrower behavior: many argue homeowners keep paying even when underwater, especially on primary residences; others list life events (job loss, divorce, rate resets, taxes) that force defaults.

Parallels to 2008 and Systemic Risk

• Some see “subprime déjà vu”: inflated book values, refusal to mark rents down, lots of empty offices, and eventual painful unwind.
• Others with risk‑management experience say regulators vs. big banks is an ongoing “game of chicken” and that post‑2008 capital/stress‑test regimes are “considerably better.”

Commercial Real Estate, RTO, and Pensions

• Thread agrees this is mainly about commercial, not residential, real estate; many loans are short term, interest‑only, with balloon payments, now hard to refinance.
• One view: push for return‑to‑office is partially to protect CRE values and, by extension, portfolios and especially underfunded government pension funds heavily invested in real estate and alternatives.
• Counter‑view: repeated skepticism of this “conspiracy”; claims that most executives push RTO based on gut feelings about culture, control, and in‑person productivity, not personal CRE exposure.

Broader Housing Anger and Zoning Debates

• Strong resentment toward “real estate always goes up,” seen as sacrificing younger generations and the “real economy” to asset holders.
• Calls to drastically increase housing supply (even “double bedrooms”) by removing exclusionary zoning, parking mandates, rent control, “character” rules, and strict preservation.
• Others argue you don’t need to literally double bedrooms; modest supply increases can have large price effects, and some cities (e.g., NYC) are already unusually dense.
• Class vs. generation: some say the problem is fundamentally class‑based; others point to data showing homeownership is indeed lower for younger cohorts at the same age.

Mortgage Structures and Leverage

• Contrast between US 30‑year fixed loans at very low rates and UK norms of shorter‑term fixes tied to prevailing rates.
• Some US borrowers treat cheap fixed mortgages as long‑term leverage, investing spare capital elsewhere at higher yields; others note banks likely hedge rate risk and that US government subsidies shape this market.

Privacy, Consent, and User Agency

• Strong backlash against enabling OneDrive folder backup without explicit, informed consent.
• Many see this as violation of user agency: even if backup is useful, it should be opt‑in or at least clearly surfaced with an easy “No.”
• Users object even if they “don’t care about privacy” per se, because the OS is doing things they didn’t ask for and making them hard to reverse.
• Concern that sensitive data (legal, medical, client, kids’ photos, regulated info) is silently exfiltrated to Microsoft’s cloud.

Risk–Benefit Views on Cloud Backup

• Some argue default backups protect typical users more from data loss (drive failure, accidental deletion) than from realistic cloud compromise.
• Others stress that “backup” here is really ongoing third‑party data transfer and not necessarily versioned or safe from account loss.
• Multiple comments highlight OneDrive’s poor implementation: high CPU, slow sync, confusing paths, and breaking workflows (e.g., Documents/Desktop moved under OneDrive\).

Microsoft Accounts, Local Accounts, and Dark Patterns

• Many try to avoid Microsoft accounts and stick to local ones, but say Windows 11 increasingly hides or blocks that path.
• Workarounds mentioned:
  • Disconnecting from the network and using a special OOBE command.
  • “Domain join instead” option on Pro.
  • Using tools like Rufus to pre‑disable online requirements.
• Conflicting reports on whether the OOBE bypass still works on latest 24H2 builds; some say yes, others say it now loops or fails.

Reversibility and Control

• Several report OneDrive silently re‑enabling or hijacking shell folders and being non‑trivial to undo (registry edits, group policy, multiple steps).
• Claims that uninstalling OneDrive “sticks” for some, while others say it reappears after major updates.

Legal, Compliance, and Regulatory Concerns

• Questions raised about GDPR, HIPAA, NDAs, trade secrets, and small practices inadvertently uploading protected data.
• Some believe this could exceed authorized access or constitute “theft,” but others note EULAs and privacy policies may try to cover it.
• Frustration that regulators (EU, FTC) are not yet aggressively addressing bundling, telemetry, and forced cloud integration.

Broader Sentiment and Alternatives

• Many see this as another step in Windows “enshittification” and a reason to move to Linux or macOS.
• Others defend the idea of default backup but condemn Microsoft’s coercive UX and ecosystem lock‑in.

Burnout, Fulfillment, and “Checking Out”

• Many describe classic burnout: arguments over architecture, politics, endless ceremonies vs. little real building.
• Some advocate “take a ticket, do the work, check out mentally” and treat SWE as a well‑paid, low‑meaning job.
• Others find this soul‑killing: they want creative challenge, ownership, and alignment with their values, not just tickets.
• A few explicitly say they can’t coast without feeling like they’re “rotting” or being fraudulent.

Team Dynamics, Ego, and Infantilization

• Frequent complaints about big egos, “one true way” engineers, and coworkers who block but don’t ship.
• Some see responsibility diffused: successes and failures are blamed on “design” or “requirements,” not individuals.
• “Infantilization” is described as perks, gamification and tone‑deaf humor instead of real autonomy, tools, or accountability.

Staying in Tech but Changing Role/Context

• Popular pivots: engineering management, TPM, project management, product design, corporate training, teaching CS, sales engineering / solutions engineering, customer‑facing consulting.
• Sales engineering is repeatedly highlighted as fitting people/strategy‑oriented ex‑SWE: technically deep, customer‑facing, good pay, but “always on” and comp below top sales reps.
• Some move to non‑profit, public sector, or “boring” but stable roles for less pressure and better alignment.

Leaving SWE for Other Paths

• Examples: manufacturing physical goods at home (laser cutters, woodworking), construction/carpentry, aerospace, acting, criminal investigation, climate‑tech investing, running small businesses (bookstore, coffee shop, dog grooming), jewelry, bookstores, travel blogging, blue‑collar “lifestyle businesses.”
• Many warn these are usually lower‑pay, higher risk, or physically taxing; often viable only with savings or a partner’s income.

Money, Risk, and FIRE

• Strong theme: nothing matches SWE for pay/comfort/accessibility; big lifestyle cuts are often required to leave.
• Some describe “retiring” in ~6 years via high comp + equity + extreme saving; others call this lottery‑like and not generally reproducible.
• Advice to save aggressively while comp is high to buy optionality later.

Sabbaticals, Mental Health, and Reassessment

• Multiple people recommend 6–12+ month sabbaticals to distinguish true career misfit from burnout.
• One commenter notes that even long breaks don’t guarantee clarity without intentional structure and goals.
• The thread includes accounts of severe distress, homelessness, and suicidality, underscoring how badly tech burnout and life crises can intertwine.

Mental health, development, and “sanitized” childhoods

• One line of argument: overly safe, idyllic childhoods create an environment mismatch with adult reality, contributing to anxiety, depression, and poor stress responses. Dark fairy tales are framed as “emotional inoculation” against later trauma.
• Pushback: many mental illnesses are seen as stemming from neurodevelopment, genetics, or direct trauma, not lack of exposure to disturbing stories. Several note that most mental illness manifests in childhood anyway.
• Some compare it to immune-system hygiene: safe, bounded exposure to “bad” things vs actual traumatic experiences, which are clearly harmful.
• There’s disagreement over whether trends like suicide rates track any of this; people cite conflicting historical data and point out confounders and reporting changes.

What counts as an “original” fairy tale?

• Multiple commenters stress that most tales are products of long oral traditions; there is no single canonical version.
• Even famous 19th‑century collections were heavily edited, repeatedly revised, and often not originally aimed at children.
• Some argue that modern simplified or happier versions are just the latest turn in this long evolution, not uniquely corrupting.

Censorship, bowdlerization, and authorial intent

• Many distinguish between:
  • New adaptations that openly rework old tales (seen as fine), and
  • Posthumous editing of existing texts while still selling them under the original author’s name (widely disliked, framed as deceptive revisionism).
• Capitalism is blamed for “lowest common denominator” versions; others note that publishers and estates also impose ideological or reputational filters.

Parenting, age, and context

• Strong disagreement over how much darkness is appropriate and at what age.
• Some parents happily read unsanitized myths and grim stories to young kids, reporting curiosity rather than trauma.
• Others emphasize developmental stages: before roughly 7, children may overgeneralize evil to “the whole world,” so heavy material can be overwhelming.
• Several stress that temperament, life circumstances, and parental guidance matter more than any fixed rule.

Cultural artifacts and alternatives

• Commenters trade recommendations: original European collections, myth anthologies, folk tales from China and elsewhere, classic children’s books, darker kids’ TV, and certain podcasts.
• Some praise specific modern shows and films (including for very young children) that handle conflict and strong emotions without gore, as examples of doing “unsanitized truth” in a developmentally sensitive way.

Overall sentiment

• Broad agreement that:
  • Children shouldn’t be lied to about the existence of evil, suffering, and death.
  • Trauma is not a legitimate “lesson.”
• Disagreement centers on whether older fairy tales are the best tools for this, and on the line between natural adaptation and dishonest erasure.

Nature of the “robot”

• Several commenters argue it’s essentially a remote-controlled rescue boat, not a “robot,” since it lacks autonomy.
• Others note many “robots” (industrial arms, bomb-disposal units, ROVs, UAVs) are also operator-controlled, so the term is defensible.
• Photos are described as looking like an RC boat with flotation strapped on, leading some to see the “robot” branding as mostly marketing.

Effectiveness vs. alternatives

• Strong support that a fast, steerable floating device can reach swimmers more quickly and safely than a human lifeguard, especially with “throw, don’t go” rescue doctrine.
• Critics suggest cheaper or simpler options: a motor + RC on a float, jetskis for lifeguards, or even “life jacket cannons.”
• Others explain why “cannons” are poor in practice: accuracy, wind, currents, difficulty donning a jacket while exhausted, and drifting devices.
• Drones that drop inflatables are proposed; examples from Australia are mentioned, though high cost is noted.

Cost, liability, and safety gear

• $12k price is seen as high; some attribute this partly to liability risk when safety products fail.
• Suggestions that similar functionality could be built much cheaper, but manufacturers face potential litigation.
• Discussion of U.S. boating laws: life jackets are generally required on board and for children to wear, but enforcement and usage are weak.

Swim skills vs. tech solutions

• Strong theme: tech is a “band-aid” unless underlying swim competence and judgment improve.
• Many recount a decline in public swimming lessons, pool availability, and institutional swim requirements (e.g., some colleges dropping swim tests).
• Others describe robust childhood programs (swim, ski, orienteering) as highly beneficial and try to replicate them for their kids.

Risk tolerance, bans, and liability culture

• Debate over pursuing “zero drownings” or “zero traffic deaths.” Some argue it leads to bans and overreach; others stress ongoing safety improvements.
• Examples: bans on unsupervised swimming, closed pools, highly controlled lakes, and large legal judgments shaping behavior.
• Tension between preventing rare tragedies and preserving access to outdoor, higher-risk activities.

Reactions to the article & writing style

• Some readers found the piece funny and refreshing, appreciating the “sleep-deprived parent encounters AI” voice.
• Others found the tone grating or distracting: Nazi jokes, “virgin/dork” jabs, and heavy snark pulled them out of the argument.
• A few remarked that this arch, mannered internet-journalist style now feels almost mass‑produced, bordering on parody of itself.
• Confusion over the intro wording (who wrote what) led some to run the article through an LLM just to extract the main points.
• Several were frustrated that the article referenced “if I didn’t have pictures you’d think I was insane” without actually showing the key pictures.

AI “slop” on Kindle and elsewhere

• Many are seeing obviously AI‑generated Kindle books, especially “safe,” saccharine, SEO‑bait children’s titles.
• Some suspect Kindle recommendations are now driven by profit over relevance, ignoring years of rich reading history and Goodreads data.
• Similar AI content floods are reported on platforms like Spotify playlists and DeviantArt; some of this has been treated as fraud and removed.
• People note Amazon is already experimenting with AI‑generated kids’ stories via Alexa.

Public perception vs tech bubble

• Several commenters say everyone they know outside tech sees generative AI primarily as spam, degraded search, worse customer service, and intrusive product features.
• They worry tech leaders are barreling ahead despite growing resentment, echoing past hype bubbles (NFTs, crypto, metaverse) but with more real-world impact.
• Others argue anti‑AI sentiment is still a vocal minority while engagement and profits stay high.

Business incentives, capitalism, and recommendations

• Strong theme: the core problem is not the models but enshittified platforms and ad‑driven, growth‑at‑all‑costs incentives.
• Recommendation systems are widely criticized as useless despite massive data collection (“you bought a printer, here are 10 more printers”).
• Some predict only large platforms and marketplaces have the incentive and capacity to seriously filter AI spam—if they don’t, they risk collapse.

Generative AI capabilities and limits

• For long-form narrative, multiple people find LLM output quickly loses coherence, repeats, and forgets prior beats.
• For short, concrete creative tasks (D&D content, setting details, fantasy recipes, cultural flavor), some report surprisingly good, useful results.
• AI is seen as decent for summaries and pattern-matching, weak for sustained originality or non-English language quality.

Indie fiction, plagiarism, and AI art

• On platforms like Royal Road, authors report their serialized web novels being scraped, lightly transformed via LLMs, and resold on Kindle Unlimited.
• Text watermarks are being tried but effectiveness is unclear.
• Many distinguish between “using AI to plagiarize books” and “using AI to generate cover art,” though some artists see both as exploitative.
• There’s extensive discussion of progression fantasy / LitRPG dominance on certain sites, driven by audience tastes and economic incentives.

User workarounds and product quality

• Some users keep Kindles in airplane mode or never connect them to Wi‑Fi to avoid ads and slop, using USB and tools like Calibre instead.
• Others say contacting support can sometimes get lock-screen ads removed.
• Broader point: many feel big consumer tech products (including Amazon’s, Meta’s, some Microsoft features) are increasingly cluttered, ad‑ridden, and degraded by unnecessary AI integrations.

Nature of gig work vs. “real jobs”

• Strong debate over whether ride-hail driving is truly flexible “day labor” or effectively a full-time job.
• Some argue drivers knowingly chose gig-style flexibility, so it’s fair for Uber to also be flexible and lock them out when demand is low.
• Others counter that many drivers want stability but have no better options; they’d prefer full-time, predictable work if it existed.
• Analogy battles: ride-hail compared to day laborers, chair-renting hair stylists, mechanics, and construction day labor – with disagreement over how comparable these really are.

NYC regulations and unintended consequences

• Central issue: NYC rules require companies to pay for drivers’ idle time, effectively guaranteeing minimum pay averaged over working time.
• Some say the lockouts are a predictable response: if Uber must pay for idle time, it must cap how many drivers can be online in low-demand periods.
• Others call the law poorly designed for not anticipating company behavior; some think it’s an intentional feature to push platforms toward true employment or fewer drivers.

Who is responsible: Uber or regulators?

• One side: Uber is just minimizing costs within the law; blame “stupid” or “antiquated” regulations.
• Other side: Uber entered the market knowing (or after) the rules; it’s their responsibility to create a good experience within those constraints, not pressure cities to change laws.
• Disagreement on whether NYC’s broader taxi regulatory regime is the real root problem.

Contractor vs employee status

• Several argue that once Uber dictates when people can work, and sets all prices, drivers function like employees and should get employee protections.
• Others insist the minimum-wage/idle-time rules are what push Uber into more employer-like control; before that, drivers could log on anytime.

Economics, exploitation, and market framing

• Some posters frame this as a stark example of labor as a commodity: surplus drivers mean low pay and poor treatment.
• Counterpoints stress voluntary participation and personal responsibility; critics respond that financial literacy and bargaining power are limited, so “choice” is constrained.

Rider experience and NYC specifics

• Complaints about being matched with non–TLC-licensed drivers who can’t legally pick up in NYC, leading to canceled rides and bad experiences.
• Debate over whether Uber could easily tech-fix this (e.g., flags for TLC vehicles) versus claiming complexity and regulation as excuses.

JSON, Observability, and Dependencies

• ntpd-rs uses JSON to expose internal state and metrics via a Unix socket; a separate Prometheus daemon and client tool consume this.
• Some question adding a JSON library for limited use, suggesting manual JSON/string construction or alternative formats (TSV, simple custom).
• Others argue hand-rolled JSON/TSV is fragile (escaping, control chars, truncation) and can cause bugs or downstream vulnerabilities.
• Debate over config formats: JSON/TOML/YAML vs minimal bespoke formats (e.g., INI-like, OpenBSD-style).
  • Pro-standard formats: fewer custom parsers, predictable syntax, reusable tooling.
  • Pro-bespoke: tiny parsers, fewer dependencies, simpler threat surface, easier long-term stability.

Performance and Timekeeping

• Several commenters care more about timekeeping quality than language choice, especially compared with chrony.
• Project contributors report ntpd-rs is close to, and sometimes better than, chrony in internal tests; algorithm docs and data are published.
• There is interest in GPS/PTS integration; project also includes a PTP implementation (statime).
• Discussion notes NTP can match or beat PTP precision given hardware timestamping and multiple time sources, but hardware typically favors PTP.

Security, Memory Safety, and Threat Model

• Some see ntpd as a low-priority target for memory-safety work; others argue it’s exactly the kind of ubiquitous, network-facing, time-critical service that merits hardening.
• NTP issues raised: runs with elevated privileges, parses bespoke binary packets, usually unauthenticated, and underpins TLS certificate validity.
• Historical C NTP implementations have had multiple CVEs; proponents argue any network boundary service should avoid C/C++ where possible.
• There is extended discussion explaining memory unsafety in C/C++ and why OS-level crashes do not guarantee safety against exploitation.

NTS and UDP Amplification

• Key risk areas: spoofable UDP, reflection/amplification, and synchronized client behavior causing unintended load spikes.
• Network Time Security (NTS, RFC 8915) is discussed as a major improvement: TLS-based key exchange plus authenticated NTP packets.
• ntpd-rs and other implementations support NTS, but adoption is described as low; many systems still use unauthenticated NTP.
• Challenges cited for NTS in public pools (key sharing, centralized NTS-KE) and ideas for certificate-based coordination.
• ntpd-rs deliberately avoids legacy management commands like monlist and limits response sizes to prevent amplification.

China, Iran, Houthis, and leverage

• Major debate over how much leverage China actually has over Iran regarding Houthi attacks in the Red Sea.
• One side: Iran is fiscally dependent on oil (claimed ~70–80% of budget), with ~90% of exports going to China, so Iran “must” listen to Beijing.
• Other side: dependence cuts both ways; China may also rely on that oil, so leverage is ambiguous and depends on alternative supply and price elasticity, which commenters say is unclear.
• Some argue Iran deliberately avoids falling under any hegemon (US or Chinese) and uses proxies (Houthis, Hezbollah, Hamas, etc.) precisely to retain flexibility.
• Another view: China currently benefits from disruption (cheaper Red Sea route and insurance for its own ships, higher freight rates it can charge, more orders for its shipyards), so has little incentive to clamp down.
• Others counter that the US, via its leverage over Israel and Gaza policy, has far more ability than China to end the Red Sea crisis but chooses not to.

---

Globalization, sea power, and “Pax Americana”

• Multiple comments connect shipping strain to a broader trend of deglobalization and waning US maritime policing.
• Some see this as validating arguments that without US (or previously UK) sea control, piracy and regional disruptions will increase.
• Others dispute the premise that the US is “retreating,” pointing to involvement in Ukraine and Gaza, though some call that support restrained or half-hearted.
• Several frame this as part of a new Cold War / BRICS-aligned struggle rather than pure economics.

---

Panama Canal mechanics and water constraints

• Many were surprised the Panama Canal relies on freshwater gravity-fed locks from inland lakes, not pumped seawater.
• Explanations: water flows from a high central lake to both oceans; each lock transit drains freshwater that must be replenished by rainfall.
• Drought directly limits transits; pumping seawater would damage ecosystems and drinking-water supplies.
• Alternatives mentioned include a Mexican interoceanic rail corridor and earlier canal concepts in Nicaragua and Colombia, but geography and politics are major barriers.

---

Shipping capacity, rates, and new entrants

• Higher freight rates prompt calls for “new entrants,” but commenters note long shipbuilding lead times, high capital risk, and incumbents’ ability to slash prices once newcomers arrive.
• Some say new, small regional carriers are appearing but lack scale to affect global rates.
• There’s disagreement on whether “an enormous amount” of new capacity is already coming online.

---

Just-in-time vs resilience

• Strong criticism of just-in-time (JIT) inventory: highly efficient but brittle; COVID-era disruptions are cited as proof.
• Some advocate “Just in Case” inventory and redundancy, but note that financial and MBA-driven cultures reward cost-cutting over robustness.
• One grocery supply-chain worker describes long delays, short product shelf life, and chronic stock issues, expressing burnout and intent to leave the field.
• Suggestions to “just buy local” are dismissed as unrealistic given current production patterns.

---

System fragility, war economies, and adaptation

• One analysis: mature global markets squeeze out all slack, increasing fragility; crises then lead to consolidation and price gouging rather than durable resilience.
• Counterpoint: extra capacity does appear, but with lag; example given is slow ramp-up of Western ammunition production after the Ukraine invasion.
• Others emphasize standard economic adjustment: current shipping crunch stems from canal disruptions, possibly opportunistic carrier pricing, and high demand; higher prices should incentivize capacity, route changes, and modest demand reduction over time.

---

Security, navies, and regional politics

• Some commenters see the Houthis’ ability to harass shipping as exposing limitations of modern navies, paralleling how Russia’s ground performance was overestimated.
• Concern that a continued US pullback from enforcing freedom of navigation could bring back widespread piracy.
• Egypt’s inaction against Houthi disruptions to Suez revenue is attributed to domestic pressure to oppose Israel rather than side against Yemeni actors.

Disaster Recovery vs. Business Continuity

• DR is described as “how do we restore IT and get back to normal?”
• BCP is “how do we keep the business running when ‘normal’ systems are unavailable, possibly without IT at all?”
• A solid BCP can keep a company alive painfully while DR is in progress; DR without BCP can still doom the business.

---

CDK, Market Structure, and Security Culture

• CDK (a dominant dealer management system vendor) is portrayed as highly entrenched with large market share alongside a small set of competitors, creating systemic fragility.
• High switching costs, deep integrations (DMS, CRM, service, compliance, networks, even printers), and retraining needs make migration very hard.
• Several comments link private‑equity style cost cutting, low engineering pay, outsourcing, and aging tech stacks to weak security and resilience.
• Others note that high-paying or “better tech” firms also suffer breaches, so compensation alone doesn’t guarantee security.

---

Operational Impact on Dealerships

• Many dealers reverted to paper: handwritten sales contracts, manual inventory walks, and ad‑hoc workarounds for F&I and service.
• Some can still sell from lot inventory; parts/service and ordering supply chains appear more disrupted.
• Experiences vary: a few report near-normal operation with delays; others describe severe constraints and lost sales.
• Thread notes the irony that a local system outage would be survivable because customers can go to another dealer, but central SaaS failure hits everyone at once.

---

Monopoly, Franchise Laws, and Dealer Model

• Strong criticism of dealer-franchise laws that block direct manufacturer sales and restrict new competing dealerships; characterized as regulatory capture and “government‑mandated” protection.
• Debate over whether independent dealers add value (test drives, warranty service, local jobs, inventory risk) versus being rent‑seeking middlemen.
• Tesla’s direct-sales model is cited both positively (transparent pricing, no “stealership” games) and negatively (service and parts delays, spotty repair networks).

---

Paper, Resilience, and Human Factors

• Multiple comments defend paper processes as essential BCP: robust to power/network failures, easy to audit, and familiar from voting systems and 911 call workflows.
• Downsides: poor searchability, physical risks (fire, loss), and degradation (e.g., thermal receipts fading).
• Cyber incidents are framed as reasons to design graceful degradation and tabletop-tested fallback processes, not to abandon digital entirely.

---

Security Responsibility and Engineering Quality

• Disagreement on blame: some say “Product” pushes insecure shipping; others argue most developers aren’t strongly advocating for security either.
• Several anecdotes from adjacent automotive SaaS describe poor testing, weak CI, minimal backup planning, and rushed, sales-driven roadmaps—seen as typical, not exceptional.

KakaoTalk’s Role and Ecosystem

• Widely described as unavoidable in South Korea; even elderly users rely on it for daily life and access to services.
• Functions as the hub of a broader “everything app” suite (KakaoTalk, Kakao T, KakaoPay, KakaoBank, Kakao Map/Metro), though not as tightly integrated as WeChat.
• Some argue its scale and service breadth are closer to Google or an Asian “super-app” than to simple messengers like WhatsApp.

Bug Bounty Policy and Security Posture

• Strong criticism that only Korean citizens are eligible for bounties, despite large foreign user impact.
• Payouts (≈$35–$7,000) are seen as extremely low given KakaoTalk’s importance.
• Several commenters argue this discourages responsible disclosure and may push researchers toward selling exploits.
• A few note such nationality limits are “normal” in Korea, and speculate about tax and regulatory complications.
• Many see the triviality of this one‑click exploit as evidence of poor security practices.

Chat App Reliability and Privacy Debates

• Broader discussion that multiple major messengers (Signal, WhatsApp, Google Chat, Telegram allegedly) have at times misdelivered messages to the wrong recipient.
• Some see this as catastrophic and inexcusable; others argue “all large software” has had serious bugs, and that complexity makes some failures inevitable.
• Long privacy debate: WhatsApp’s E2E encryption vs Meta’s broader surveillance history; Telegram’s non‑default E2E but more trusted leadership for some; strong preference by a subset for open‑source, decentralized alternatives.

Legal and Ethical Issues Around Security Tools

• Question whether releasing exploitation tooling would be illegal in Germany under §202c StGB.
• One side claims mere possession of such tools can be criminal; others respond the law hinges on intent, and that research and self‑testing aren’t prohibited.
• Concern that vague cybercrime laws become “three‑felonies‑a‑day” tools of selective enforcement.

Korean Tech, Protectionism, and “Sovereign Software”

• Discussion of Korea’s parallel tech ecosystem (Kakao, Naver, LINE, local ride‑hailing) and limited penetration of US giants.
• Supporters frame this as data sovereignty, job creation, and protection from US platform power.
• Critics argue protectionism and government picking winners lead to technically weak, insecure products (e.g., long‑mandated ActiveX).
• Debate over whether American dominance is due to superior products or geopolitical “colonization” of the web.

Work Culture and Startup Environment

• Some attribute security shortcuts to hierarchical culture, non‑negotiable deadlines, and focus on visible features over invisible security.
• Others counter that similar hierarchies exist at large Western firms; the difference is chaebol and government dominance crowding out independent startup influence.
• Korean startups receive significant government grants but face conservative VC, heavy paperwork, and incentives that can make them quasi‑state‑driven.

Localization, UX, and Foreign User Experience

• Multiple reports that Kakao‑based taxi and bank services are hard for foreigners: Korean‑centric language, bank‑account requirements, limited support for non‑residents.
• Debate over whether it’s reasonable to expect high‑quality English and other languages in apps aimed primarily at domestic users.
• Some argue multi‑language support is standard and good business; others say foreign tourists are too small a market to justify prioritization.
• KakaoTalk itself is localized into many languages, but supporting apps (Kakao T, KakaoMap, banking) are seen as less consistently localized and more difficult to use.

Perceived Microsoft Strategy

• Many see a long-term push toward “Windows as a subscription service,” with everything metered (Office 365, Copilot, Xbox Game Pass, Outlook ads).
• Windows 11’s pressure to use a Microsoft account is viewed as part of a broader push for logged‑in identities, telemetry, upselling, and ad targeting.
• Some argue Microsoft no longer mainly cares about install base, but about authenticated users tied to cloud services.

Microsoft Account Requirement & Workarounds

• Users note you can still create local accounts using tricks: disconnecting from the internet, using OOBE\BYPASSNRO, selecting domain join on Pro/Enterprise, or entering a “banned” email.
• These methods are seen as fragile “Konami codes” that Microsoft could disable at any time.
• Some report Home/Pro setups that refused to proceed without internet, but relented on captive portals (e.g., paywalled Wi‑Fi).

Comparisons: macOS

• Debate over whether macOS is “basically unusable” without an Apple ID; several report fully usable Macs with no account, losing only iCloud and some services.
• Apple is criticized for hardware lock‑in, expensive storage, non‑replaceable components, planned obsolescence, and telemetry.
• Others still see macOS as far less user‑hostile than Windows, especially on ads, accounts, and surveillance.

Linux and BSD as Alternatives

• Many are moving to Linux (Debian, Fedora, Mint, Pop!_OS, Zorin, Fedora Silverblue, Bazzite, etc.) or even GhostBSD, often triggered by Windows 11, Recall, and telemetry.
• Opinions diverge on beginner‑friendliness: Debian seen as principled but hard; Mint, Pop!_OS, Zorin praised as more “just works” for ex‑Windows users.
• OEM preinstalls of Ubuntu/Linux exist but are perceived as niche, especially in business.

Gaming: Windows vs Proton

• For several, Windows remains only for gaming, anti‑cheat, and some VR/AR (e.g., Roblox, WMR, certain multiplayer titles).
• Others report near‑full satisfaction gaming on Linux via Steam + Proton and Heroic, with most major titles working, though:
  • Some see ~10% performance loss and occasional stutters.
  • Anti‑cheat DRM and non‑Steam launchers remain problem areas.

Windows LTSC / IoT as Escape Hatch

• Windows 10/11 LTSC (and IoT Enterprise) are praised as “de‑bloated” builds: fewer ads, less telemetry, less feature churn, and still good gaming compatibility.
• Licensing is a major obstacle: officially volume/enterprise‑only. Many discuss GitHub activation scripts and local KMS emulators, openly acknowledging this is piracy but rationalizing it if they already own a consumer license.
• Some fear Microsoft could tighten LTSC to herd users back into mainstream, account‑centric SKUs.

Broader Sentiment

• Strong sense that Windows has become adversarial software requiring constant “whack‑a‑mole” against ads, telemetry, and unwanted features.
• Several users express a principled decision to stop supporting companies that show “disdain” for users, seeing migration to Linux/Mac as a moral as well as practical choice.

Adoption & real‑world experience

• Some participants ask if anyone runs _FORTIFY_SOURCE=3 in production and whether it has caught real bugs or caused noticeable slowdowns.
• Arch Linux reportedly builds with level 3; this encourages some to try it.
• Debian has long used _FORTIFY_SOURCE=2, suggesting level 2 is considered mature.

Performance impact & missing data

• The article’s claim that “gains outweigh the cost” is criticized as unsupported.
• Commenters want concrete performance and code size measurements, not just assertions that costs “may be worth it.”

Crashes, “false positives,” and production behavior

• Concern that fortified checks convert some previously “working” (though unsafe) patterns into crashes, increasing non‑exploitable failures.
• Some argue this is good: a loud crash at the overrun point is better than silent corruption or later mysterious aborts.

realloc, malloc_usable_size, and compatibility

• Fortification and GCC’s dynamic object size logic interact badly with patterns using realloc but continuing to use old pointers.
• This ties into pointer provenance: two pointers with the same address may not be interchangeable; using the old one after realloc is UB.
• Debate over malloc_usable_size: one side says the behavior of using “excess” bytes used to be allowed in practice and got retroactively declared buggy; others counter it was always “diagnostic only” and changing man pages doesn’t break existing binaries.

UB, strict aliasing, and compiler flags

• Long sub‑thread on how aggressive optimizations driven by undefined behavior (integer overflow, strict aliasing, pointer rules) surprise programmers and can delete safety checks.
• Some compensate by always using flags like -fwrapv and -fno-strict-aliasing, valuing predictability over maximal speed.
• Others argue these rules exist for important optimizations and that C/C++ are inherently performance‑oriented languages.

Tooling, documentation, and ecosystem

• One commenter links an OpenSSF compiler‑hardening guide as a broader reference.
• Another wants better documentation of how GCC hardening features and sanitizers interface with non‑glibc, freestanding/nolibc environments.
• Sub‑discussion about difficulty contributing to large projects (GCC, glibc, LLVM) and the need for maintainers to engage more clearly with outside contributors.

Meta / Miscellaneous

• Minor discussion about HN title formatting and article “newness” given it’s from 2022.

Dialogues and narrative gimmicks

• Mixed reactions to inline dialogues with fictional characters in technical writing.
• Some find them engaging when they anticipate reader objections; others see them as distracting, hard to follow, and more for the author’s amusement than the reader’s.
• Simpler Q&A or FAQ-style asides are often preferred. Good execution is seen as rare and high-skill.

Progress indicators vs scrollbars

• Debate over page progress bars: fans say they help when pages have long footers, comments, or infinite scroll where the native scrollbar misleads.
• Critics find them redundant, visually noisy, or “re-implementing the scrollbar badly.”
• Underlying complaint: modern OS/browser scrollbars are tiny or auto-hidden, pushing sites to reinvent them.

Tables of contents, headings, and anchors

• Strong support for ToCs on long/technical posts; seen as overkill or distracting on short/essay-style pieces.
• Several patterns mentioned: side ToC that highlights current section (“scroll spy”), collapsible ToC, or dedicated ToC pages.
• Many want every heading to be linkable, with small anchor icons and/or copy-to-clipboard; mobile-friendly CSS patterns are discussed.

Footnotes, sidenotes, and asides

• Highly divisive: some love rich sidenotes and footnotes (citing specific blogs as inspiration), others see them as clutter or evidence of undisciplined writing.
• Mobile handling is a recurring problem; some favor details/summary blocks or pure footnotes that always work in narrow layouts and RSS.

RSS, dates, and longevity

• Many argue RSS is still widely used and essential, not a “microfeature” or dead technology. Others feel it’s neglected or in slow decline.
• Strong consensus that visible publish/update dates are crucial context; anger at blogs that hide them for SEO or marketing reasons.
• Some use multiple dates (start, first publish, last modified); URL-embedded dates can confuse readers without explanation.

Link previews and external markers

• Interest in per-site link icons and hover previews, but also concern about privacy (unwanted requests), mobile usability, and “UI land mines.”
• Some like previews only when they genuinely save a click by summarizing the target.

Other appreciated or disliked microfeatures

• Liked: plain-text versions, good print CSS/PDFs, helpful 404s with suggestions, single-page indices of all posts, blogrolls/webrings, “start here” and favorites pages, client-side search, comments bridged from the fediverse.
• Disliked: hijacking keyboard arrows for navigation, sticky headers/footers, auto-open comments, left-aligned narrow columns on wide screens, justified text on the web, and JS-heavy “overengineered” blogs that hurt performance, especially on mobile.
• Several argue many of these features should be browser-level (ToC from headings, better scroll indicators) rather than bespoke per site.

EU decision and DMA context

• EU Commission press release says Apple’s current App Store terms breach the Digital Markets Act (DMA), especially anti‑steering rules and its “core technology fee”.
• Key preliminary findings: developers cannot freely tell users about cheaper alternatives; “link‑out” steering is heavily constrained; Apple charges fees even for purchases made on the web shortly after a link‑out.
• There’s a pending EU court case over whether the DMA can mandate “free of charge” access to iOS APIs, which some see as central to judging Apple’s new fees.

Anti‑steering rules and fairness

• Many argue it’s fundamentally wrong that Apple forbids developers from:
  • Saying Apple takes a percentage.
  • Stating that subscriptions are cheaper on the web.
  • Seamlessly steering users to external payments.
• Others note anti‑steering clauses are common in platforms (Amazon, Airbnb, card networks) and see Apple as behaving like other intermediaries, not uniquely bad.

Competition, monopoly, and lock‑in

• Pro‑regulation comments frame Apple as a de facto gatekeeper with:
  • Strong ecosystem lock‑in (especially in the US via iMessage/social pressure).
  • Power to tax and veto entire business models on iOS.
• Counter‑arguments:
  • Apple’s device share is smaller than Android’s; users can choose Android.
  • Console stores (Sony, Nintendo, Microsoft) and Steam also take ~30%.
• Several insist the real issue is not 30% itself, but that on iOS there is:
  • Only one store.
  • No practical way to distribute apps outside Apple’s control.

Security, user protection, and choice

• Supporters of Apple’s model emphasize:
  • Centralized billing, refunds, and scam protection.
  • Comfort that only Apple holds card data.
  • Reduced malware risk for less technical users (e.g., older relatives).
• Critics reply:
  • Laws and chargebacks already deter fraud.
  • “Security” is also used to justify anti‑competitive restrictions and blocking web/PWA capabilities.
  • Sophisticated users should be free to sideload or use alternative stores without affecting default safety for others.

Economic impact and possible outcomes

• Debate over whether Apple might exit the EU:
  • Some see it as possible but unlikely given EU’s revenue share and network effects.
  • Others note DMA fines (up to 10–20% of global revenue) could outweigh EU profits.
• Many expect “malicious compliance”: Apple doing the legal minimum, delaying features (e.g., AI, screen mirroring) in the EU, and testing the boundaries through courts.

CentOS 7 EOL and CentOS Stream Lifecycle

• CentOS Linux 7 support ends imminently; some users treated it as effectively dead earlier due to repo/update issues.
• CentOS Stream 8 is already EOL; CentOS Stream generally has ~5‑year lifecycles tied to the “full support” phase of the matching RHEL release.
• Some users who picked Stream to avoid the CentOS 8 early EOL were surprised to learn Stream 8 is already out of support.

Migration Paths and Tools

• Suggested technical paths: LEAPP (RHEL/CentOS Stream 7→8/9), convert2rhel (CentOS→RHEL), AlmaLinux ELevate (CentOS 7→Alma 8/9 via intermediate steps).
• Reports are mixed: some say upgrades are smooth; others call convert2rhel/LEAPP “a nightmare,” with repeated failures and broken PHP, Python, mail, and DB setups.
• Workarounds include rebuilding from scratch, using containers (run EL7 userspace on newer hosts), or temporarily relying on extended support services.

Replacement Distros & Ecosystem Fragmentation

• Common replacements: AlmaLinux, Rocky Linux, RHEL proper (including the “free for up to 16 systems” program), Oracle Linux, Debian, Ubuntu Server.
• Alma/Rocky generally reviewed positively; Alma tracks CentOS Stream, Rocky aims for bug‑for‑bug RHEL compatibility.
• Some move entirely to Debian/Ubuntu, often citing simpler upgrades, lack of corporate “shenanigans,” or dislike of Ubuntu’s CLI/log “ads” and snaps.
• SUSE’s Liberty Linux Lite is noted as a commercial lifeline for CentOS 7 via repo swap, but its minimum spend is seen as high.

Enterprise Support, Compliance, and Trust

• For FedRAMP/FIPS and similar regimes, CentOS (especially Stream) is problematic; RHEL 8/9 and specific minor streams (e.g., 8.6, 9.2) are actively kept compliant.
• Debate over relying on free rebuilds vs. vendor promises: some distrust IBM/Red Hat’s policy changes; others distrust Oracle more, despite Oracle Linux’s long, consistent availability and fast security updates.
• Extended support options mentioned: CIQ, OpenLogic, TuxCare, Red Hat’s own RHEL7 offerings, and Debian ELTS via Freexian.

Upgrade Pain and Technical Issues

• Large CentOS 7 estates (HPC clusters, scientific environments, on‑prem VMs, VPS fleets) report major effort and risk, especially with highly customized “pet” systems.
• Specific issues: secure boot with CentOS Stream 9, network interface naming bugs in leapp‑upgrade, missing OpenSSL 1.1 headers on RHEL 9, floating‑point differences between CentOS 7 vs. Rocky 8 setups.
• Some argue major version bumps are inherently disruptive; others note that decade‑long ABI/API stability in enterprise distros makes big, infrequent jumps unavoidable.

Background: 2021 Deal & Alleged Breach

• Thread centers on Boeing allegedly breaching a 2021 deferred prosecution agreement (DPA) over 737 MAX fraud.
• Some see the $2.5B settlement + no prosecution as effectively a “bribe” to avoid criminal liability; others say it is a standard fine/settlement mechanism, not personal enrichment.
• One commenter notes key missing detail: what exactly Boeing failed to do under the DPA (compliance overhaul, reporting, honesty) is still unclear.

Criminal Liability: Company vs Individuals

• Strong sentiment that prosecuting only the corporation is inadequate; many want specific executives and managers charged.
• Debate on how far down to go: C‑suite only, or also engineers, test pilots, line workers who signed off on unsafe work.
• Counterpoint: workers were often under pressure and retaliated against for raising safety concerns.

Nature of the Crime: Fraud vs Manslaughter/Murder

• Legal framing is fraud against the FAA, not homicide. Several argue murder charges are unrealistic but manslaughter or corporate manslaughter (as in UK law) would be conceptually appropriate.
• Others argue the deliberate profit‑driven concealment of safety issues is morally close to murder.

Regulation, Capture & Systemic Issues

• Widespread view that FAA and DOJ have been too lenient, enabling repeated safety failures (MAX crashes, door plug blowout, counterfeit titanium, missing parts).
• Seen as a symptom of regulatory capture and a broader erosion of “rules-based order,” where breaking rules is normalized and lightly punished.

Remedies Proposed

• Ideas include:
  • Massive fines “into bankruptcy,” charter revocation, or a corporate “death penalty.”
  • Temporary or partial nationalization, forced board/CEO purge, or court‑appointed overseers.
  • Breaking up Boeing or at least separating military and commercial units.
• Others argue expropriation would clash with U.S. constitutional protections and is politically unrealistic.

Markets, Shareholders & Incentives

• Many blame shareholder‑value focus and executive stock-based pay for safety shortcuts.
• Debate over whether punishing shareholders (through large fines or expropriation) meaningfully disciplines management, given principal–agent problems and passive index investing.

International Comparisons & Safety Concerns

• China is cited as harsher on corporate malfeasance (including executions and long sentences) and as the first regulator to ground the MAX.
• Some fear further crashes could severely damage public trust in air travel; a few say they already avoid Boeing flights.

Cross‑platform and architecture support

• Cosmopolitan (Cosmo) binaries can be “fat” APE executables that run on both amd64 and arm64, including modern macOS on Apple Silicon.
• The same binary can run across multiple OSes (Linux, macOS, FreeBSD, Windows, etc.), sometimes as a native binary, sometimes via being parsed first as a shell script.
• There are edge cases: some Linux distros (e.g., CentOS/Rocky 8, certain ARM64 kernel configs) still have issues; a fix for older address-space assumptions helped but didn’t solve all cases.
• Android kernel support exists at the binary level; app-level integration is unclear.

Mechanics, POSIX changes, and APE polyglots

• APE binaries are polyglot: valid DOS/PE/ELF/Mach‑O plus shell script headers.
• POSIX shell rules were relaxed to allow binary data after an initial shell-parsable region, enabling this trick.
• There’s debate whether POSIX “approves” the specific implementation; some argue the spec text doesn’t fully sanctify APE-style payloads.

Use cases and projects

• Notable uses: llamafile (portable LLM runner), redbean (single-file web server with Lua), and demos like portable Python and GNU tools.
• WAMR (a WASM runtime) has Cosmo support, enabling higher-level stacks like Hermit.
• People are interested in portable GUI stacks via embedded web UIs, Java/Graal Native, and potential Python single-file tools.

Performance, libc design, and comparisons

• Cosmo focuses on fast, static, cross‑OS binaries and a self-hosting toolchain (cosmocc).
• Claims: 2× faster than musl for many CLI workloads due to vectorized string routines; fast malloc with per-core arenas for threaded code.
• Some benchmarks (e.g., rsync on Windows) show slow real-world performance, but others attribute this to Windows I/O rather than Cosmo itself.
• The C++ STL is being “reinvented” to reduce include bloat and drastically improve compile times.

Skepticism, stability, and standards conformance

• Enthusiasts see Cosmo as a way to avoid heavy runtimes (JVM, Node, Electron) and to ship “download once, run anywhere” binaries.
• Skeptics worry about:
  • Reliance on loader/ABI corner cases that may break on future OS releases.
  • Non‑conforming libc behavior vs POSIX/glibc.
  • Long‑term production suitability.
• Others counter that Cosmo now avoids many brittle tricks (e.g., MAP_FIXED), depends on stable ABIs, and aims for longevity.

Alternatives and open questions

• Comparisons to WebAssembly: Cosmo uses native CPU code (faster) vs VM bytecode; WASM still attractive for some.
• Questions remain about GPU support, broad binfmt integration across Linux distros, and whether Cosmo will see widespread mainstream adoption.