On Jan 9, 2008, at 9:34 AM, Wichmann, Mats D wrote:

>
>> Yep. Sorry for not mentioning. There's likely some slightly better
>> means than AutoReqProv: no that should be done, but AutoReqProv
>> is one way to disable dependency extraction.
>
> yes, well, as long as the dependency extraction leads to package
> names, which have no guarantee of being consistent across distros
> (and I have *lots* of examples of where they arent), it does need
> installing.
>

The issue is scriptlet interpreter dependencies iirc. but you've
likely greased
	Requires: /bin/sh
into some version of the LSB packaging spec.

The polluted name space (i.e. different distros have different
package names for identical content) is very real however,
and the only engineering solution that I can think of is
mapping names from one space to another. The infrastructure
to maintain the mapping transforms is non-trivial no matter what.

>
>> Better would be to insturment custom lsb-find-provides and lsb-find-
>> requires with actual tests of, say, permitted file paths and content,
>> and emit     Requires: lsb iff the tests pass. E.g. many FHS tests
>> could easily be done automagically on every build.
>
> yes, this would be nice.
>

Make a list of tests that might be appropriate to automagically
check during build please.

>> No matter what, all this stoopid configgery to produce a LSB
>> "standard" package could be simplified to a --lsb option in rpm if
>> anyone cared to ask me to implement.
>>
>> No one has.
>
> we talked about that once, so I'd say I have
>

Yep, we've talked. No worry, --lsb is going into rpmbuild Real Soon Now.

>> But I do ask the question:
>>     What use is rpm as a LSB "standard" format if everything in
>>     rpmbuild needs disabling? Isn't cpio or tar or ... a better
>> choice than a lobotomized rpmbuild?
>
> some people seem to think so, but they're archive formats, not
> packaging formats - if "payload only" is okay, then by all means
> (and some developers do prefer that).
>

The difference between archivers and package managers is metadata.

If one chooses to forbid all rpm metadata that cannot be represented
in, say, tar archives (with conventions like "lsb-" naming instead of  
RPMTAG_NAME),
then of course there is no difference between "payload only" archive and
rpm package formats.

Methinks that was the intent with LSB "standard" packages. I  
certainly remember
a year's worth of phone calls discussing how tar archives could be  
used as
alternative LSB packages way back when. And (iirc) tarballs _ARE_ the
recommended containers for source packages, not *.src/.rpm packages.

73 de Jeff