HNHacker News
TopNewBestAskShowJobs

hobofan

8,360 karma · joined March 21, 2014

Jack of all trades (- master of some?)

goisser94@gmail.com

https://hobofan.com

Mastodon: @hobofan@toot.berlin (Twitter: @hobofan)

---

Currently building https://github.com/EratoLab/erato

submissionscomments
hobofan··on GitLost: We Tricked GitHub's AI Agent into Leaking Private Repos
> If you configure GitHub to allow public code or LLM instructions to run in contexts that have access to sensitive things, they will leak; that’s not GitHub’s fault, it’s yours.

Is there a way to segment access per agentic workflow, so that you can have both habe an agentic workflow that has access to sensitive data and one that has only access to public data? Is the default to set the scope to only the current repository? Does Github appropriately inform about the risk of combining an agentic workflow with access to private repository data?

If the answer to any of those questions is "no", then that's a problem.

(Classic GH Workflows are also riddled with priveledge escalation via PR-triggered workflows, but that's another topic.)

hobofan··on Road to Elm 1.0
> I mean Elm has removed features breaking compat

To be fair, Elm hasn't made it to 1.0 (yet). That's where languages should make breaking changes before being stuck with the flaws forever.

hobofan··on Road to Elm 1.0
I've head the displeasure of working with more than one Elm zealot in the past, and also allowed a service to make it into production (which was a huge disaster). Due to that, I know a lot more about Elm than I'd like to.

I think the "Elm is stable not dead" seen from the few people that stuck around with Elm is largely cope for being stuck with an unmaintained language. Languages, like all other pieces of software need maintenance or they degrade in the world moving around it (e.g. there is/was no official aarch64 build of Elm in the period of non-maintenance).

I also would say that Elm is still largely unfit for most realistic production scenarios, unless you have the manpower to build everything from scratch, as interoperability with the outside web world (JS/TS) is an afterthough, and by some parts of the community not desired.

hobofan··on Road to Elm 1.0
The project was dead. The previous release was 7 years ago, where it stopped because the creator (same as author of this announcement) stopped maintaining it and since the community hasn't progress beyond a BDFL-model that's where it died.

So it was dead, it just now has been resurrected (and AFAIK with a whithered community in the meantime).

hobofan··on The bottleneck might be the air in the room
HEPA-filter air purifier and a robot vacuum that is scheduled to run while your are not in the apartment (to reduce baseline dust) are probably the most simple/cost-effective measures.
hobofan··on Half-Baked Product
Seeking approval. It's apparent from everything he posts on X + the memeing + the whole "I am a elite Path of Exile"-saga.
hobofan··on Looking Ahead to Postgres 19
That is not a homebrew sytax that Postgres came up with themselves.

It's SQL/PGQ, which was derived from the Cypher language for Neo4J and now is part of the SQL standard.

hobofan··on Open Source Low Tech
I think it makes more sense if it's not viewed under the aspect of trying to be maximally productive in its output. If you look at it over they years, they've also tried to do a share of public outreach / education (though I'm not sure how successful), and I think they also sustain themselves in terms of labor with burnt out tech people that are looking for a change of scenery.

So similar to a good zoo, that does both active conservation work, and at the same time public education (e.g. in the form of guided tours for school trips).

> I guess that's why it seems to languish in obscurity.

I think even in a well-executed form, it would likely still be quite obscure, as there is next to no need for it in western societies (apart from emergency preperation).

hobofan··on South Korea to spend $1T on more memory chip production and humanoid robots
I don't have the historic numbers at hand, but I would assume that for each of those categories this is a similar proportional increase, so it's similarly notable to mention the increase in spending on humanoid robots.
hobofan··on We all depend on open source. We will defend it together
That's just your typical list that makes up the Linux foundation.

It might not be the idealistic flavour of open source you prefer, but it's the flavour of open source that's actively in use in most tech companies, and that also forms the makeup of most corporate open source participation (e.g. also the top corporate Linux contributors).

hobofan··on Minimus container images are now free
> 2. Isn't there a slight risk of upstream attacks being amplified by this?

I think the argument would be that consuming Minimus' containers would have a less severe amplification (or even reduction), as all upstream attacks that rely on a combination of third-party vulnerabilities would be rendered infeasible (since they reduce the amount of third-party dependencies in an image).

> 3. This ties into 2, if someone downloads and uses an image that is later found to be compromised they mostly have no way of being notified that happened.

For this you need a consumption-aware scanner anyways (e.g. that lists images running in your Kubernetes). Anything else will be too spammy, as you can't notify for everything for you have at some point in time have used as a base image.

hobofan··on Minimus container images are now free
Asking the very obvious question (as it's not apparent from the website): Why would I use this over DHI (Docker Hardened Images) or Chainguard Images, both of which also have a set of free hardened images?
hobofan··on United Wizards of the Coast recognized by NLRB
There are ~25k total different MTG cards. Who cares if even 1k of them are for collabs (as long as they don't contain broken cards that end up in tournament play)?
hobofan··on Trains halted across Germany because of communication system problem
Probably someone forgot to renew the TLS certificate.
hobofan··on SMPTE Makes Its Standards Freely Accessible
Does that actually mean "freely" (as in beer) available, or it must be able for anyone to be procurred?

I would assume that France, like many countries heavily rely on ISO standards, which in many cases cost some money. Or is it the case that those standards are not explicitly mandated, but are essentially only fulfillable by knowing the ISO standard?

hobofan··on Google workspace threatening to block Firefox access
Organization admins may roll out hardened Firefox settings via their MDM solution, and then based on that want to restrict usage to Firefox.
hobofan··on Zero-Touch OAuth for MCP
For organizations it's definitely an advandtage, as it also allows for preventing data flowing to personal accounts. Right now, any employee could knowingly or unknowingly exfiltrate data, by just connecting to e.g. the personal instead of the organization's JIRA account, and there is essentially no standard way to guard against that as an organization.

That's also blind spot in normal OAuth and I was surprised to find that this was never addressed in all the years that non-MCP OAuth found adoption.

hobofan··on Zero-Touch OAuth for MCP
Just to be clear, enforcing proper access control and data seperation is the job of a client (/harness), not the job of an LLM Provider (though of course the most popular clients are the first party clients for the LLM providers).

Ensuring that an LLM doesn't have free reign over calling any MCP tool at any point in time is one of the main jobs of a client (apart from the general data persistence, etc.), and one that's very dependent on the setup (e.g. many MCP servers expose public data where tool calling is mostly not that sensitive) and the acceptable risk profile.

This MCP extension also doesn't significantly change anything about tool calling control from the perspective of the client. MCP servers were previously also authenticated once per user, and not once per conversation or once per tool call.

hobofan··on Zero-Touch OAuth for MCP
I don't think you are missing anything obvious. Entra ID doesn't support DCR, and the state of the ecosystem here is suboptimal.

The typical way to do MCP OAuth is with traditional up-front registered clients. However, in practice a lot of MCP clients work with an assumption that DCR works, and as you point out don't offer an option to specify a client ID.

However, some clients do support that (ad: our tool Erato does[0]), and the typical solutions deployed in enterprises do, where MCP access is usually centralized via a web UI like ours. One alternative that also exists is are MCP gateways, which do pre-registred oauth between the gateway and service, and allow for DCR between the gateway and clients.

[0]: https://erato.chat/docs

hobofan··on How we run Firecracker VMs inside EC2 and start browsers in less than 1s
I'm a bit surprised that with all this, they still stuck with Chromium.

We have a much less sophisticated setup in our web-access MCP server[0] where browser instances are spawned as subprocesses and the biggest win in stability, CPU and memory usage we had was in switching from Chrome to Lightpanda[1].

Fitting to the statement at the end of the article, the faster browser to boot might be one that allocates less memory in general.

[0]: https://github.com/EratoLab/web-access-mcp

[1]: https://lightpanda.io

hobofan··on Only 16 Percent of Americans Think AI Will Have a Positive Impact on Society
Most optimistic outcome: Hardware & software advances keep pace; Open weights & locally runnable models keep pace with frontier (at least the open weights part seems to hold), making AI advances widely available to anyone; Individual productivity skyrockets, having knock-on effects in most industries, slowly allowing less and less people to do the work required to maintain the basic needs of our civilization(s); UBI utopia achieved.

Do I think that's likely? No, but mostly because it would require parallel uprising of the global working class, which doesn't seem to materialize.

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
I'm not sure what official communication you are seeing but e.g.[0][1] it was clearly stated that it would fall under that. And according to their framework of "game must stay in a reasonable playable state", of course the online mode of PEAK, Battelfield, Call of Duty - which are all games that are primarily purchased for their online mode (but do have an offline mode / singleplayer campaign) - would qualify for their proposal.

[0]: https://youtu.be/sEVBiN5SKuA?t=1967

[1]: https://youtu.be/36qDEeTDXNE?t=2378 - Statement by affiliated Pirate Party member

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
> How is LAN susceptible to DDoS? I think you're thinking of P2P, unless you're worried about your LAN buddies trolling you I guess.

I didn't say that. Games don't to P2P anymore because it's susceptible to DDOS. LAN as alternative to Internet P2P may have seperate drawbacks (but may obviously disregard DDOS protection).

> The legislation is being discussed largely because of what EA, Activision, and Ubisoft do to games with millions of paying customers. Letting the hardest edge case set the ceiling for consumer protection is a convenient outcome for the people least affected by it.

Then propose some legislation that actually deals with that and doesn't in its communication ignore the cost of retrofitting games for all developers of games with any online component. But that's not how SKG riles up it's base of support.

> The implication that informed consumers are a problem for the industry is a pretty damning admission about how the industry currently operates.

This doesn't have anything to do with informed consumers / uninformed consumers. My comment was about the unwillingness of customers to pay for monthly subscriptions.

> consumers are a problem for the industry

They honestly are. My main reason for not going into game development is the incredible entitlement from the customerbase. I understand that situations like Control or The Crew suck, but those should be individual class action lawsuits. But instead everyone wants to impose legislation on essentially all online games, even those they purchase for less money than a coffee at Starbucks. And all that just to apparently be fine with buggy, laggy, borderline playable versions of the game, with worse matchmaking UX.

That things you bought stop working is not unique to video games with online components. I can run very few games that I bought as a child, because I don't have a CD-ROM drive or a Windows 98 machine anymore.

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
Yes, you are right, but that's not the part in discussion here with GP.
hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
All hundred different microservices? And I'm sure you also want docs for how to run them all together too?
hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
> Besides, PEAK has offline mode already built in, so no subscriptions needed.

Yes, single player offline. SKG is specifically about deactivating the online/multiplayer component in games, so if PEAK were to deactivate their online coop mode, they would definitely fall under it (ignoring grandfathering).

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
> PEAK uses P2P networking as far as I can tell.

PEAK is listed as an example of using the Photon Fusion Shared Authority model[0]. AFAIK Photon Fusion doesn't allow for any proper P2P in any of their network topologies (apart from the fully offline single-player mode).

[0]: https://doc.photonengine.com/photon/v1/photon-products#the-q...

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
> WoW, FF14, Elder Scrolls: Online, runescape

There are in total ~15-30mil paying subscriber that pay subscription fees to a single game. That's 1% of all gamers globally.

But sure, let's slap a $5/month subscription fee on all upcoming games that would have the chance to be the next Among Us / PEAK, and let's see how well that works out for indie developers.

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
> > you can just make your game into a service requiring a monthly payment

>

> Ah, yes the simple option of "completely tank your playerbase".

And gamers at large shouldn't pretend that they are going to be shelling out money for subscriptions. There is a reason that even most MMOs switched away from mandatory subscription pricing (apart from the outlier of WOW), and it's not to make the publishers filthy reach, but often barely viable.

hobofan··on Stop Killing Games fails to secure EU law despite 1.3M signatures
ARMA has famously awful netcode.

Minecraft has mandatory hosted servers that are so non-trivial to host that it spawned a $XXmillion industry of third-party hosting, and first party hosting is also a paid service. Also bad netcode in the Java edition.

I'm not sure if I'd call either of those "just fine". That system also only works for games that are able to attract some minimum viable community where some enthusiasts volunteer to host any servers at all. Great for a really long-lived community, bad for any new game trying to make a splash.

← PreviousPage 8 of 34Next →