Apache handling of .txt.gz

Apache handling of .txt.gz

[Gerald Pfeifer]

Hi overseers,

here is some issue Geoff noticed with our Apache config; does any
of you have an idea how to tackle this?

(Thanks, Geoff, for analysing this!)

Gerald

> all my .txt.gz files were reporting
>
> Content-Type: application/x-gzip
> Content-Encoding: x-gzip
>
> which is wrong (that would mean the file is compressed twice), it should be
>
> Content-Type: text/plain
> Content-Encoding: x-gzip
>
> I think this is because /usr/share/magic.mime has:
>
> 0       string          \037\213        application/x-gzip
>
> I worked around this by putting stuff in a .htaccess file, but thought
> you should know...

Re: Apache handling of .txt.gz

[Jason Molenda]

On Wed, Feb 05, 2003 at 05:31:44PM +0100, Gerald Pfeifer wrote:

> >
> > I worked around this by putting stuff in a .htaccess file, but thought
> > you should know...

What did you put in the .htacces file?

We run a non-standard module, mod_gzip, on sourceware.  If you ask for
index.html and an index.html.gz is also present, and the client browser
claims it can decompres gzip-encoded content, the .gz version is sent.
My initial reaction is to wonder if mod_gzip might be getting involved
somehow.

J

Re: Apache handling of .txt.gz

[Christopher Faylor]

On Wed, Feb 05, 2003 at 10:12:26AM -0800, Jason Molenda wrote:
>On Wed, Feb 05, 2003 at 05:31:44PM +0100, Gerald Pfeifer wrote:
>
>> >
>> > I worked around this by putting stuff in a .htaccess file, but thought
>> > you should know...
>
>What did you put in the .htacces file?
>
>We run a non-standard module, mod_gzip, on sourceware.  If you ask for
>index.html and an index.html.gz is also present, and the client browser
>claims it can decompres gzip-encoded content, the .gz version is sent.
>My initial reaction is to wonder if mod_gzip might be getting involved
>somehow.

I haven't loaded mod_gzip since we upgraded sourceware.  I've been sort
of wondering if someone would notice and complain.  I suspect that this
is related.

cgf

Re: Apache handling of .txt.gz

[Geoff Keating]

> Date: Wed, 5 Feb 2003 10:12:26 -0800
> From: Jason Molenda <jason-gcclist@molenda.com>
> Cc: overseers@gcc.gnu.org, Geoff Keating <geoffk@geoffk.org>,
>    Janis Johnson <janis187@us.ibm.com>

> On Wed, Feb 05, 2003 at 05:31:44PM +0100, Gerald Pfeifer wrote:
> 
> > >
> > > I worked around this by putting stuff in a .htaccess file, but thought
> > > you should know...
> 
> What did you put in the .htacces file?

I put:

# This is wrong; something is deciding that .gz indicates
# a type as well as an encoding, and RemoveType doesn't seem to fix
# it.
AddType text/plain .log .sum .gz

[There's also other stuff, but this is the line that fixed the problem.]

> We run a non-standard module, mod_gzip, on sourceware.  If you ask for
> index.html and an index.html.gz is also present, and the client browser
> claims it can decompres gzip-encoded content, the .gz version is sent.
> My initial reaction is to wonder if mod_gzip might be getting involved
> somehow.

It's possible.  I thought that the problem was in
/usr/share/magic.mime because there are two magic files, and they differ;
/etc/httpd/conf/magic has:

0       string          \037\213        application/octet-stream   x-gzip

but /usr/share/magic.mime has:

0       string          \037\213        application/x-gzip

and if I have my columns lined up right the second one looks like it
might be causing the problem.

-- 
- Geoffrey Keating <geoffk@geoffk.org>

Re: Apache handling of .txt.gz

[Jason Molenda]

On Wed, Feb 05, 2003 at 01:17:54PM -0500, Christopher Faylor wrote:

> 
> I haven't loaded mod_gzip since we upgraded sourceware.  I've been sort
> of wondering if someone would notice and complain.  I suspect that this
> is related.
> 

The fact that mod_gzip isn't loaded leads me to believe it's not related,
but I haven't really looked at all.  Geoff's explanation is as likely as
anything I can think of.

The lack of mod_gzip isn't something people will notice unless they
perpetually bring up old mailing list index pages or documentation,
and even then they're more likely to assume the net is just slow.

For instance, http://gcc.gnu.org/ml/gcc/2003-01/index.html is 370kbytes.
http://gcc.gnu.org/ml/gcc/2003-01/index.html.gz is 40kbytes.  
The end user will never realize that the page loaded 10x faster
than it really should have when they get the .gz, but users over
slow links are going to be a lot happier getting the compressed
version.

The same thing is done for all the documentation in gcc.gnu.org/onlinedocs
(and maybe the same for gdb's as well).


J

Re: Apache handling of .txt.gz

[Gerald Pfeifer]

On Wed, 5 Feb 2003, Jason Molenda wrote:
> but users over slow links are going to be a lot happier getting the
> compressed version.

At home I have ISDN (8KB/s), and having mod_gzip really would be nice
(if it doesn't interfere with other stuff too much).

Gerald
-- 
Gerald "Jerry"   pfeifer@dbai.tuwien.ac.at   http://www.pfeifer.com/gerald/