> From: Joe Orton [mailto:joe@manyfish.co.uk]
> Sent: 16 February 2002 16:08
> On Fri, Feb 15, 2002 at 11:46:40AM +0100, Sander Striker wrote:
>> The problem we saw previously appeared to be on the server side.
>> Thanks to Justins patch, this seems to work fine now (I did a
>> full checkout without problems).
>
> Does mod_deflate now stream responses to the client (i.e. isn't always
> precomputing the Content-Length)?
Not yet. We dropped the original(false) Content-Length though, so that
fixes the problem. We can do streaming later.
>> This patch will enable subversion to handle compressed responses
>> from the server. As Joe pointed out, the util.c patch could be
>> left out, since that gives compressed OPTIONS responses. I'm
>> not completely familiar with how it all ties together, but are
>> the reports fetched in fetch.c or in util.c? If the latter,
>> I think it makes sense to leave the compression in.
>
> Yes, good point. I found a potential segfault in neon's
> ne_decompress_destroy, and I'm also just hooking up code to check the
> gzip response checksum: it might be prudent to wait until there is a new
> neon release with these in before enabling this by default.
Ah, I noticed that commit to neon, yes. So I think we should archive
this patch until the new neon version is available.
> Regards,
>
> joe
Thanks,
Sander
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:37:08 2006