[PATCH] native inline gravatar

Andy Green andy at warmcat.com
Wed Jul 4 02:44:18 CEST 2018

On 07/04/2018 08:34 AM, Jason A. Donenfeld wrote:
> On Wed, Jul 4, 2018 at 2:28 AM Andy Green <andy at warmcat.com> wrote:
>> I looked at it, but there's no md5 api in JS... you have to do it by
>> hand in JS.  It's possible but I think it might be slow if it hits a
>> page of 50 different email addresses.
> MD5 is really fast and fits easily into javascript's integer types.
> There should be reasonably fast implementations available.
> Alternatively, you could generate asm.js and web assembly code using
> something like emscripten.
>> Is it supported to do the job of pygments in Lua?
> You can do any type of text processing you want from Lua. If you'd
> like to contribute a Lua highlighter, that'd be quite nice.
>> What it does today is spawn the python runtime, start that up etc.
> Have you measured latency issues with this?

Nope, but it's obviously more efficient to not be spawning python apps.

Don't you think?

>> I know... bandwidth and server compute time.
>> Also there's no reason (other than not enough care taken with frees atm)
>> cgit should only be buildable as a cgi starting up its own process each
>> time.
> You're talking about FastCGI support? I have no idea how this is
> related to the current line of discussion -- perhaps just by being
> another gripe or something -- but this is actually something we've

What was the first gripe?  I am providing patches for all these things 
not moaning about them.

I mention it because starting up cgit as a process each time is also 
overhead, I think it would be interesting also to look at eliminating that.

> looked at in depth. (See the list archives.) Our last conclusion from
> examining it was that so much of libgit is not re-entrant, and so we'd
> need to move to something like libgit2 for this to be feasible. Too
> many globals, etc.

Let me guess, not enough people around working on this project, for some 
reason, to move to libgit2.

>> It could optionally link against OpenSSL then :-)  It can optionally
>> link against lua.
>> If you're not interested in going where this stuff is going, you can
>> save us both a lot of time by just saying it now, and I'll stop trying
>> to sell it here.
> I made that clear in my very first message to you:
> "Sorry, but not a chance something like this can be accepted."

If that's your feeling about the whole stack of clientside stuff, I 
understand you don't want to cooperate on it.

No worries I'll leave you to it then.


More information about the CGit mailing list