On Dec 13, 2007, at 2:25 PM, Ralf S. Engelschall wrote:

> On Thu, Dec 13, 2007, Jeff Johnson wrote:
>
>> Hmmm, what are you asking for?
>>
>> Here's the current behavior with a "Blah: this should fail" tag in  
>> a spec
>> file:
>>     $ xxxrpm -ba time.spec
>>     error: line 19: Unknown tag: Blah: this should fail
>>     $ rpm -E '%{_arbitrary_tags}'
>>     Foo:Bar:Baz
>>
>> My guess is that you want the "Tag_0x12345678" generic syntax  
>> hardened.
>
> No, no, the implemented behaviour is exactly what I want. Only the
> explicitly listed tags are allowed. So, I'm fine with the  
> implementation
> status-quo. But I thought I remember that *you* wanted to allow even
> *unknown* (i.e. not explicitly configured) tags and I was surprised  
> this
> evening to discover that this is not what got actually implemented.  
> So,
> from my point of view: KEEP THE IMPLEMENTATION AS IS. But I was under
> the impression that *YOU* originally wanted something different. Or  
> I at
> least misunderstood your original intentions ;-)
>

Hehe, ok. No I was trying to precipitate a workable definition for  
"arbitrary"
this summer, playing Devil's advocate. Once I had a workable definition,
I knew what had to be implemented.

(aside) I often argue out of both sides of my mouth while trying to  
figger out
what to do. Some issues, like whether rpm should carry an explicit
arbitrary tag string dictionary in all *.rpm packages, or whether  
arbitrary tags can be
used as a virus infection route, are quite subtle.

Re-read all my "Berkeley DB haters" postings with that additional
perspective too. I'm quite serious about attempting a "flat file" rpmdb
while snacking on *.deb packages in rpm-5., and lord knows everyone
wants SQL, just not the current schema please.

73 de Jeff