XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

LWN.net comments

This feed contains the text of all comments posted to the LWN.net site.

Internet protocols

Now, this does remind me of something... now that we talk about internet protocols. Something I have been thinking about for a while now. Or at least for a couple of days. UDP is a protocol that is designed particulary for data transfer. It's especially designed for data transfer. Not for communication. But, anyways, you couldn't trust a hacker with anything they can get their hands on. UDP is commonly used for communication on the internet. What an absolute hack! *namaste*

Succinct syntax versus verbosity

I was looking at a C library and its documentation saying "the return value of this function is allocated, free it" versus "the return value of this function is part of X other structure, don't free it", and those return values had the same type The GTK/GNOME project (a lot of C code) has developed GObject Introspection which adds an annotation system written in special API comments, for what the C syntax lacks. This is used for automatic language bindings to make the C library usable for a…

Scrollback?

I have this binding: bind-key -n F2 run-shell 'tmux new-window "urlview <(tmux capture-pane -p -t $( tmux display -p "#{pane_id}" ) -S - -J)"' which takes the entire history of the current pane and plumbs it into urlview in a new window so that I can open URLs easily. I'm sure that one could use this to put it all into a file (and an editor to help trim) or do things like trim around "start of command" and "end of prompt" embeddings.

AI-drop project (AGPL)

Hi! > I get now, that your real proposal is kind of a federation to collect bad IPs and distribute it between peers or operate in some kind of subscriber model with an external "upstream", correct?" Correct! > How the identification/honeypot works is unrelated to that, e.g. the group concept you described is not part of the linked repository. There is no code published which actually operate with these groups, correct? Not yet, but the plan is to add some code people can put on honeypots and…

Meh

Are you sure you're not misunderstanding what that camp wants? (I may misunderstand it, though.) I think what they want to say may be that using contracts and relying on them to throw is a valid approach to modify the control flow of the program. They're not saying that you shouldn't be able to remove it from your compiler any more than they're saying that you cannot have a Python interpreter that never raises exceptions. Just you get what you deserve if you do that.

The SCTP flaw

I'm speaking of the Linux kernel SCTP flaw. Nobody has a fix for that yet because Xint disclosed it before it got merged into any maintainer's tree. Trying to keep the nginx thing under wraps is probably just going to make things worse because embargoes no longer work when the fix is out, but instead of Xint blogging about the SCTP LPE they should've sent a v2 of their patch so it can get merged.

The SCTP flaw

Debian still doesn't have its fix out; this clearly wasn't coordinated very well if at all. I suppose the fact that it was independently discovered at least three, if not four times is a solid enough reason to push the disclosure early (the CVE thanks three groups for responsible disclosure, but Rapid7 also released a writeup about them discovering it today).

Scrollback?

I'm with you. I'm sure there is a reason for it but the scroll back thing in screen and tmux is a bit wank. I can use two finger scrolling on the touchpad to scroll within a terminal. I've just tried it in screen/tmux and atuin (history) rocked up instead! From memory, without atuin, you end up scrolling back within the terminal instead of within screen/tmux. Sadly, that means I end up using tmux/screen simply to ensure that something important finishes running, if I remember to use them. I've…

The SCTP flaw

And yep, my "this just feels irresponsible" has been confirmed: they sent a fix[1] to the mailing list (easily identified, since they gave us the release the SCTP flaw was introduced in and there are only 44 non-merge commits to net/sctp/ in that) but did not follow up with a v2 after a maintainer review on the 12th of November. Instead, they went on a publicity tour. Nice! [1]: https://lore.kernel.org/all/20261110224233.82977-1-yves.b...

The SCTP flaw

Did they really need to go public with this before someone wrote a fix for the SCTP flaw, just to showcase a full chain to root? nginx RCE on its own would've been enough to make people care. This just feels irresponsible.

Exit ISO

the twist is that what you call the lobbyist group includes the creator of C++: <https://www.stroustrup.com/Contracts_fundamental_flaws.pdf> In a BDFL model, contract would not stand a chance.

Any idea of a new Julia LTS release?

Yes, I've seen the official word. I was wondering if somebody had some "insider" info whether there is some plan to eventually make a new LTS release, or is that considered some kind of failed experiment that will eventually fade away whenever the current LTS ages out of support.

Re: Bribery

Ignoring the comment is probably the best thing that could have happened. "Not worthy of our attention." As far as the project(s) themselves are concerned, 200,000+ lines of code is not "easily" human reviewable by a single person as a single logical construct. It can only be reviewed based on smaller, discrete packets of information. That's part of the problem with long standing small & large projects running into various logical fallacies once reviewed by automated methods. From the…

Maybe we should just give the whole "internet" thing a break...

If we turn the internet off, all those agents will stop working! Even the people with in-house data centers won't be able to file any bug reports online; they'll need to meet up with each other to file their reports, and then when the internet is turned back on, we'll have all the updates ready! More seriously, it looks like the 'most severe' of the attacks requires sending 128 gigabytes of data; so with luck, we won't see this one be seriously exploited over the internet.

Biggies incoming

And have you seen what's coming next? Greg has just posted the next batch of stable -rc kernels. We have 2,651 patches for v7.3, and a dizzying 3,318 for v6.6 - no wonder Greg is burned out. But how stable can we really consider the stable kernels, with such an absolutely vast inflow of changes, even to a kernel that came out 3 years ago?

Succinct syntax versus verbosity

If one part of your code uses reference counting, that doesn't mean everything does. If you're designing C code, and have cross-reference pointers from one structure into another, then reference counting is a really good idea there, too. If you're not using reference counting, you end up with complex rules for "this is the owner of the structure" versus "this pointer doesn't own the target, but here's how we make sure the structure stays alive as long as the pointer exists", and down that path…

Any idea of a new Julia LTS release?

The official word is: “You should only be using the long-term support (LTS) version of Julia if you work at an organization where implementing or certifying upgrades is prohibitively expensive and there is no need for new language features or packages.” Almost everyone should use the current stable release (1.13 at the moment).

AI-drop project (AGPL)

I am blocking 1.000.000 right now. But I blocked 4.000.000 as well. And I really see that it works on rocketgit.com. I see. Maybe I just gave up to early (70.000). I get now, that your real proposal is kind of a federation to collect bad IPs and distribute it between peers or operate in some kind of subscriber model with an external "upstream", correct? How the identification/honeypot works is unrelated to that, e.g. the group concept you described is not part of the linked repository. There is…

Scrollback?

Ah, the dreaded alternate vs. normal screen buffer. Does adding the scrollback-related entries in https://github.com/geertu/board-fram/blob/main/examples/s... to your .screenrc work for you?