httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Behlendorf <>
Subject Re: Signatures, and sealing wax, and..
Date Tue, 12 Aug 1997 19:19:28 GMT
At 05:58 AM 8/12/97 -0400, Rodent of Unusual Size wrote:
>    As I recall, for 1.2.0 we had some disparities when it came to the
>    content of binary kits.  I'd like to suggest the following, and open
>    the ball for discussion.  For 1.2.3, or whenever we next do a
>    release with binaries.  Wherever I say "should" below, I mean that
>    all the binaries must follow the same rules, whatever we decide them
>    to be.
>    1. Binaries should be built with the vanilla Configuration (i.e.,
>       no additional modules).  Possible exceptions: mod_status and
>       mod_info?

Those two, plus dbm_auth, anon_auth, expires, headers, and setenvif.  The
last one in particular since we have to provide lines to make sure broken
browsers still work.

>    2. httpd image should include platform name (maybe
>       "httpd.`helpers/GuessOS`"?)

As always...

>    3. Tar files should be available compressed with compress(1) *and*
>       gzip(1) (nothing new here).

I'd like to dump "compress", personally.

>    4. Compressed tarchives should have accompanying .md5 *and* .asc
>       (PGP) signature files available.  (Yes, Ben, I know it's not as
>       good as signing the uncompressed tarchive, but it means people
>       can verify what they copy from the site w/o having to uncompress
>       it first.)

As mentioned, not all binary builders will have access to PGP or an MD5
tool.  Though those do exist on hyperreal, so anyone with an account
uploading their own builds to there (or handling one from a trusted source)
could handle that.  Making is a suggestion and not a requirement is preferred.

>    5. src/Configuration should use the platform's native cc(1) if it's
>       considered good, and *not* gcc - unless the native cc is suspect
>       or downright broken (HP-UX, can you hear me calling? ;-).

This should be up to the binary builder, in my opinion.

>    6. README files should all be virtually identical, describing the
>       above and whatever minimal exceptions there may be.


Should we recommend a level of optimization used?  I.e. if gcc, use -O2?


"Why not?" - TL  - -

View raw message