[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: aptitude --target target sarge xxx doesn't limit to sarge

On Mon, Mar 06, 2006 at 06:15:28PM -0500, Daniel B. wrote:
> Paul E Condon wrote:
> >On Mon, Mar 06, 2006 at 12:47:45PM -0500, Daniel B. wrote:
> >
> >>Paul E Condon wrote:
> ...
> >>>I think apt does not recognize "sarge" or "woody" or "etch" etc.
> >>
> >>... why?  ...
> ...
> >I think it is a design flaw, more than an implementation bug. 
> Why does it seem to be a design flaw?  (What makes it seem to be
> bigger, or harder to fix, than just an implementation bug?)
> > Speaking
> >from the outside of Debian, it seems that the 'design' of apt grew
> >over time.  Whether 'stable' is an alias for 'sarge' or 'sarge' is an
> >alias for 'stable' is a big deal in a 'well designed' system, but this
> >issue seems not to have been resolved as apt was implemented. Now to
> >'fix' it would involve major rewrite ...
> What about APT suggests that the rewrite would be major?

Look at the structure of the directories and the contents of the files
in a Debian repository. Distributions are organized by code name in some
places and by status name in others. Metadata files contain both code
names and status names. The working programs expect this and it is
hard coded into them. The repository is a database, but it does not
have what RDB people call 'data independence'. Historically, the transition
of a code name distribution was fraught with difficulty for all involved.
The introduction of the pool helped a lot, but the system remains fragile.
I love Debian, but the repository system, files, directories, and software
is very complicated. To change the way distribution names are handles by
*all* the software consistently will be a lot of work and nearly
impossible to do either in one fell swoop, or gradually either.

Just my opinion. I may be wrong. You can file a bug, if you want.
Once I realized that the theory sections of the documentation are
misleading, and only certain usages are given in the examples, I
looked at the organization and decided the way it is is way beyond
being fixed with a few intelligent suggestion by me, and nevertheless,
good enough for me.

If all the metadata about packages were kept in a proper relational
database then all the various special purpose programs for answering
various questions about what is in the repository could be replaced
with SQL queries. The code-name/status-name issue would be recognized
in the normalization and this issue would be resolved with finality.
But it won't happen just by filing a bug, IMHO.

Paul E Condon           

Reply to: