Bug #32274 for HTTP-Async: [PATCH] Make connect and write async as well

# Sun Jan 13 07:03:49 2008 mbarbon [...] users.sourceforge.net - Ticket created

Subject:

[PATCH] Make connect and write async as well

Download (untitled) / with headers
text/plain 641b

Hi, I noticed that in HTTP::Async, connect and write are blocking operations. Since I need to handle gracefully network problems and big POST to slow servers, I performed some changes to HTTP::Async. I believe most of it could be done in a subclass if needed, but since it seems a useful functionality, I am submitting here as a patch. Apart from tests for the functionality, it adds: - A Net::HTTP::NB subclass to handler asynchronous connect and buffered write - adds a second IO::Select object to HTTP::Async to handle writes, plus the necessary code to handle buffered writes Let me know what you think. Regards, Mattia

Subject:

darcs.diff

Download darcs.diff
text/x-diff 12.7k

Message body is not shown because it is too large.

# Sun Feb 10 05:09:07 2008 my.tranquil [...] gmail.com - Correspondence added

From:

my.tranquil [...] gmail.com

Download (untitled) / with headers
text/plain 1k

On Sun Jan 13 07:03:49 2008, MBARBON wrote: Show quoted text

> Hi, > I noticed that in HTTP::Async, connect and write are > blocking operations. Since I need to handle gracefully > network problems and big POST to slow servers, I performed > some changes to HTTP::Async. > > I believe most of it could be done in a subclass if needed, > but since it seems a useful functionality, I am submitting here > as a patch. > > Apart from tests for the functionality, it adds: > - A Net::HTTP::NB subclass to handler asynchronous connect and > buffered write > - adds a second IO::Select object to HTTP::Async to handle writes, > plus the necessary code to handle buffered writes > > Let me know what you think. > > Regards, > Mattia

Hi, i think it's similar to the problem i've encountered. i've posted a new bug on how the timeout param isn't transfered, so that way the fact the connect is syncronicly can really damange the async way of things in the module. can you send me the new module you've rewritten, (NBR), do you think of submitting it to CPAN? thanks, eyal

# Sun Feb 10 05:09:09 2008 The RT System itself - Status changed from 'new' to 'open'

# Mon Feb 11 04:12:51 2008 evdb [...] ecclestoad.co.uk - Correspondence added

Subject:	Re: [rt.cpan.org #32274] [PATCH] Make connect and write async as well
Date:	Mon, 11 Feb 2008 09:12:16 +0000
To:	bug-HTTP-Async [...] rt.cpan.org
From:	"Edmund von der Burg" <evdb [...] ecclestoad.co.uk>

Download (untitled) / with headers
text/plain 1.8k

Hello, I'm really pleased that you guys are using HTTP::Async and have changes you want to make to it. Please feel free to grab the latest code from the repo: http://dev.ecclestoad.co.uk/svn/cpan/HTTP-Async/ http://dev.ecclestoad.co.uk/trac/cpan/browser/HTTP-Async and either send me some patches or the output of 'htpasswd -n' and I'll give you a commit bit. As I have very little time at the moment the commit bit would probably be a better bet. There are some changes in there since the latest one on CPAN. Cheers, Edmund. On 10/02/2008, eyal edri via RT <bug-HTTP-Async@rt.cpan.org> wrote: Show quoted text

> > Queue: HTTP-Async > Ticket <URL: http://rt.cpan.org/Ticket/Display.html?id=32274 > > > On Sun Jan 13 07:03:49 2008, MBARBON wrote:

> > Hi, > > I noticed that in HTTP::Async, connect and write are > > blocking operations. Since I need to handle gracefully > > network problems and big POST to slow servers, I performed > > some changes to HTTP::Async. > > > > I believe most of it could be done in a subclass if needed, > > but since it seems a useful functionality, I am submitting here > > as a patch. > > > > Apart from tests for the functionality, it adds: > > - A Net::HTTP::NB subclass to handler asynchronous connect and > > buffered write > > - adds a second IO::Select object to HTTP::Async to handle writes, > > plus the necessary code to handle buffered writes > > > > Let me know what you think. > > > > Regards, > > Mattia

> > Hi, > > i think it's similar to the problem i've encountered. > i've posted a new bug on how the timeout param isn't transfered, so that > way the fact the connect is syncronicly can really damange the async > way of things in the module. > > can you send me the new module you've rewritten, (NBR), > do you think of submitting it to CPAN? > > thanks, > > eyal > > >

-- evdb@ecclestoad.co.uk - http://ecclestoad.co.uk