IRC channel logs


back to list of logs

<amz3`>so my js-closure function doesn't properly capture the closure
<amz3`>that the last bug I need to tackle before being able to run my framework
<amz3`>and that is a success!
<amz3`>I found the bug, fixed and my framework works
<amz3`>and it's not slow, yet ;)
<amz3`>which means I will have a nice week end
<happy_gnu[m]>amz3`: great :)
<amz3`>the code is here
<ijp>amz3`: yeah, that one is easy done
<amz3`>indeed, it was not hard if i succeed
<amz3`>still a win!
<dustyweb>beep hello
<dustyweb>I need to implement HTTP Signatures for Guile
<dustyweb>because it turns out!
<dustyweb>this is being implemented by all ActivityPub implementations in practice
<dustyweb>this is good because HTTP Signatures is a solid spec!
<dustyweb>the bad news is I can't use Guile's built-in web server
<dustyweb>reason being that the way we handle headers is "nice" in that everything's ported to actual types you can mess with
<dustyweb>however, with http signatures when we read a request we're going to have to quasi-canonicalize it
<dustyweb>and unforutnately
<dustyweb>that involves preserving the whitespace etc of headers
<dustyweb>and also the ordering!
<dustyweb>ordering, I think we can do that, but whitespace, not the case.
<dustyweb>so I'm thinking of what I have to do, and best I can tell I should do the following:
<dustyweb>read the headers for canonicalization while also writing the original contents out to a string port
<dustyweb>and *then* pass the string port's contents over to guile's normal headers parsing code
<dustyweb>ramble ramble ramble
<dustyweb>anyway! that's what's up with me
<civodul>dustyweb: oh so these are signatures computed over the set of HTTP headers?
<civodul>we had a similar "problem" for narinfo signatures FWIW, where we have to keep the exact initial narinfo representation (order, whitespace, etc.) to be able to verify the signature
<dustyweb>civodul: signtures of both headers and body, but headers need to be able to be "normalized", yeah
<civodul>so the signature is over normalized headers, or over the headers as-is?
<dustyweb>semi-normalized, but now I'm confusing myself!
<dustyweb>normalized at least in terms of "they are sorted"
<dustyweb>at the very least we'll hit the problem that this needs to be able to normalize headers guile doesn't "know about"
<civodul>if it's "normalized" it's less bad than if it needs to be as-is
<civodul>perhaps Guile could do it
<civodul>after all it already does a pass of "normalization"
<dustyweb>yeah maybe
<dustyweb>civodul: though would still need to handle terms that Guile currently "drops"
<civodul>hmm yeah
<daviid>amz3: you could ask the libgit2 team to add en entry for guile-git (actually send a patch ...)
<amz3`>daviid: a patch about what?
<amz3`>daviid: an entry in guile projects?
<daviid>about their front page and bindings page ...
<amz3`>daviid: oh
<daviid>so guile-git appeas on the front page, next to chicken-git :) ...
<amz3`>daviid: yeah, but it's not complete yet
<amz3`>daviid: I mean it's far from complete even if useable
<amz3`>daviid: sure I will submit a PR for that once I find it good enough
<amz3`>I was supposed to work on hat this week end
<daviid>not usable? I thought guix was using it already
<amz3`>yes, it's useable but incomplete
<amz3`>but since guile-js is working... I think I will give guile-js a go ;)
<daviid>ah ok