I don't use CRM114, but I just read Christian Perrier blog post on
planet.debian.org so I happened to check out this bug report.

Milan Zamazal wrote:
> This is not the first time .css format change has happened.  And yes, it
> was reported and discussed upstream in the past.  They are willing to
> avoid such changes but it may not always be easy.  More pressure
> (i.e. from more users) on the upstream maintainers may or may not
> improve things.  The .css format is generally not considered as very
> fixed, e.g. it's not portable across architectures after all.
> 
> There were other upgrade problems in the past.  For instance, I
> personally don't simply believe .crm scripts from the last version work
> with a new one without changes and I always test that crm114 still works
> after the upgrade.  I'd say that backward compatibility is not a crm114
> feature and one must expect problems when upgrading crm114 to a new
> version.  Of course, a user may not think about crm114 when typing
> `apt-get upgrade'.  This wouldn't be a big problem if crm114 wasn't
> typically used for things like e-mail delivery.
> 
> So the primary problem is how to _warn_ users that crm114 gets upgraded
> with higher than casual risk of breaking things.  The current practice
> of warning about important changes in NEWS.Debian seemed to work well so
> far.  But I agree it may not be enough and I'm open to suggestions how
> to improve it.

If upstream can't keep the .css format stable, then they should break
it more obviously so that crm114 can fail gracefully.

They should have a version number in the .css file header.  If they
find the wrong version number in a .css file, then crm114 should exit
immediately with a nonzero exit code, to indicate to the caller that
it can't handle mail, and the caller should hold on to the message in
whatever way it knows how (for example, if fetchmail is calling
crm114, then fetchmail wouldn't mark the email as read, and wouldn't
delete it from the server, and so that fetchmail could download the
same message in the next run.)

Just my two cents.

--Ken

-- 
Chanoch (Ken) Bloom. PhD candidate. Linguistic Cognition Laboratory.
Department of Computer Science. Illinois Institute of Technology.
http://www.iit.edu/~kbloom1/

Attachment: signature.asc
Description: Digital signature

Reply via email to