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

Re: Design Document: Version 0.01



Hi,
>>"Jason" == Jason Gunthorpe <jgg@gpu.srv.ualberta.ca> writes:

Jason> On 30 Sep 1997, Manoj Srivastava wrote:

>> If a package file contains foo version 2, and another package file
>> contains foo version 3, and foo version 1 is installed, then
>> irrespective of the order of reading package files I should get foo
>> upgraded from version 1 to version 3. This is not likely to happen
>> often, so invoking the version camparison routine should not slow
>> things down much.

Jason> That is all up to the UI. Upgrading is done selectively based
Jason> on what distribution (stable/unstable) and user choice in that
Jason> matter.

	Quite. Given that the distributions are selected as one I want
 to upgrade from, there is a default that is set up (dslect shows a
 list of upgraded packages). I am talking about the default upgrade
 option offered: I want it to be very easy to upgrade to foo version 3.

	I am, I guess, questioning the requirement for listing the
 Packages files in a given order. I think that if package foo is
 insatalled, it should be automatically be offered for upgrading,
 unless there is a hold on it. If there are more than one new versions
 in the Packages files, the most recent should be taken as the default
 to upgrade to (recent <=> highest version).

	Is there any good reason not to do so? If the same version of
 foo is available in two packages files (hence two locations), one may
 be chosen at random (Hmm, cost of upgrading would be nice to use in
 this matter: cheaper to upgrade from disk than from ftp server)


Jason> The cache can store N versions of the same package,
Jason> each with distinct depends, provides, etc.

	Sounds good.

Jason> As part of the cache generation all version lists are sorted
Jason> (no duplicates are permitted) which means the only time
Jason> versions are compared is when depends are being analysed.

	Umm, does that mean that the previous capability is useless
 since no more than one version shall be entered? (that does not sound
 right. I must be missing something. Where's the coffee?)

	I mean, we have to enter two versions if one is installed and
 a new version is also present, right?

	manoj

-- 
 Machines should work.  People should think.  -- IBM motto
Manoj Srivastava               <url:mailto:srivasta@acm.org>
Mobile, Alabama USA            <url:http://www.datasync.com/%7Esrivasta/>


Reply to: