[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Use of versioning in archiving.
From: |
Richard Frith-Macdonald |
Subject: |
Re: Use of versioning in archiving. |
Date: |
Tue, 4 Dec 2001 09:27:05 +0000 |
On Tuesday, December 4, 2001, at 06:13 AM, Gerrit Van Dyk wrote:
Hi all,
I logged a bug yesterday in connection with the unarchiving of NSValue.
Thanks Richard for responding so quickly. I was just wondering now, I
saw
that the encoding of NSValue also changed without using versioning. Is
this
a common practice within GNUstep to change some of the encoding methods
without checking for versioning?
No.
In fact, the 'final' version of my changes included versioning,
moderate optimisation, and a few other fixes and tidyups.
Our system makes use of Blob's in the
database, these blobs are created using archiving and then unarchived
when
their details are needed. If GNUstep does not check for versioning then
our
blobs will become unusable the moment the archiving methods change.
Should I
start using different ways of archiving our objects into blobs, just to
make
sure that our system will not become unstable with a new release of
GNUstep.
I can also see a potential problem over DO if the versions of GNUstep on
either side of the connection is not 100% the same.
Well ... it shouldn't happen. As of GNUstep-base 1.0, newer releases of
the
code should remain capable of decoding archives produced by older
releases
(at least within a major release version).
However, mistakes/bugs will obviously creep in on occasions ...
fortunately,
with GNUstep you are better off than with a proprietory system ... if an
incompatible coding change gets in to a release, you can get quicker help
and/or fix the problem yourself and contribute the fix back.