librelist archives

Re: [mongrel2] Merged in >2G Support, Please Check my git-fu

From:

Dalton Barreto

Date:

2011-06-03 @ 23:56

2011/6/3 Zed A. Shaw <zedshaw@zedshaw.com>:
> On Thu, Jun 02, 2011 at 10:12:01PM -0300, Dalton Barreto wrote:
>> 2011/6/2 Zed A. Shaw <zedshaw@zedshaw.com>:
>>
>> > And, test out the >2G support.  What this merge does is send the file as
>> > 1G sized chunked-encoding pieces until it's done.
>> >
>>
>> I was thinking about this implementation. HTTP chunked responses is
>> better suited when we don't know the size of the response. So my idea
>> is to change this implementation to just write chunks to the IOBuf,
>> but not HTTP chunked encoded.
>
> Ah, the *only* problem with that is that there's a very good chance the
> receiving side can't handle a content-length that big.

Actually maybe this is already happening, as for a static file with
1.9GB mongrel2 is
responding with a response this big. Do you really think that this
could be a problem?

> My thinking with
> the chunked-encoding is that it will work without having to set a giant
> content-length header at the beginning.
>
> I'd say leave it as-is, it's not much of a difference and has some
> advantages.  I also refactored it so take a look to see how I improved
> it some.

Alright. I didn't see this mail and actually I committed the
modification to remove the http chunked encoding logic. But if http
chunked is really better I can revert the changes, no problem.

Alex tested the implementation on a OSX and it was not working, did
you tested it yourself on a OSX? did it worked?

Let me know and I will revert the changes I made today. And sorry for
the "commit noise".

Thanks!

-- 
Dalton Barreto
http://daltonmatos.wordpress.com
http://wsgid.com

On Fri, Jun 03, 2011 at 08:56:10PM -0300, Dalton Barreto wrote:
> Alright. I didn't see this mail and actually I committed the
> modification to remove the http chunked encoding logic. But if http
> chunked is really better I can revert the changes, no problem.

Well, leave it without chunked-encoding then and then I'll try both and
see what I like best.

> Alex tested the implementation on a OSX and it was not working, did
> you tested it yourself on a OSX? did it worked?

I only tested it on linux.

-- 
Zed A. Shaw
http://zedshaw.com/

2011/6/4 Zed A. Shaw <zedshaw@zedshaw.com>:
> On Fri, Jun 03, 2011 at 08:56:10PM -0300, Dalton Barreto wrote:
>> Alright. I didn't see this mail and actually I committed the
>> modification to remove the http chunked encoding logic. But if http
>> chunked is really better I can revert the changes, no problem.
>
> Well, leave it without chunked-encoding then and then I'll try both and
> see what I like best.

Alright. Maybe we will never find out a client that does not handle
big responses and this implementation will be just fine. Let's see.

>
>> Alex tested the implementation on a OSX and it was not working, did
>> you tested it yourself on a OSX? did it worked?
>
> I only tested it on linux.
>

Ok.

-- 
Dalton Barreto
http://daltonmatos.wordpress.com
http://wsgid.com

>
>
> > Alex tested the implementation on a OSX and it was not working, did
> > you tested it yourself on a OSX? did it worked?
>
> I only tested it on linux.


Yeah it's screwed up.  bsd_specific's sendfile has been needing some TLC for
a while, so I'll do that and see if I can't sort things out on the OS X side
tonight.

2011/6/4 Alex Gartrell <agartrell@cmu.edu>:
>>
>> > Alex tested the implementation on a OSX and it was not working, did
>> > you tested it yourself on a OSX? did it worked?
>>
>> I only tested it on linux.
>
>
> Yeah it's screwed up.  bsd_specific's sendfile has been needing some TLC for
> a while, so I'll do that and see if I can't sort things out on the OS X side
> tonight.

Great, Alex. Let us know when you change bsd's sendfile so we can test
the chunked responses implementation and see if it starts working.

Thanks

-- 
Dalton Barreto
http://daltonmatos.wordpress.com
http://wsgid.com