Loading pieces of the code would be nice but most people load everything and then take what they need to add just that. If there are tests that fail we should address those issues. Until there is a framework for loading parts of packages I think this would make things harder to use and understand then allowing people to get a working version and then remove what they don't want for their own deployment. Also consider the overlap. We have a number of hashes that use the same streaming framework. It would be easier to load cryptography and then delete SHA1 then it would be to package up all the components needed for MD5 and then package all the same components again for SHA256. The code base is not so large, that the benefit would be worth the time it would take and confusion it would cause.
Ron Teitelbaum
-----Original Message----- From: cryptography-bounces@lists.squeakfoundation.org [mailto:cryptography-bounces@lists.squeakfoundation.org] On Behalf Of Germán Arduino Sent: Tuesday, May 10, 2011 7:21 PM To: The general-purpose Squeak developers list Cc: Cryptography Team Development List Subject: [Cryptography Team] Re: [squeak-dev] Re: Some question regarding portability
2011/5/10 Sven Van Caekenberghe sven@beta9.be:
On 10 May 2011, at 14:45, Andreas Raab wrote:
- Are announcements available ? Can this framework be loaded ?
What is the official way to do this ?
There are at least four different implementations of the
announcements framework. They are not (fully) compatible with each other. The question is: which one is required by Zinc?
I am not aware of that many versions. I subclass Announcement, instanciate an Announcer, send #announce:
and #on:send:do to it, that's all.
Then how about adding ZnAnnouncement, ZnAnnouncer with the two or
three methods that you actually need? Announcements always struck me as more of a pattern than a framework.
I agree that Announcements are pretty simple. Still, if it is available, then why not use it. Sharing and reusing general purpose classes seems pretty important and
elementary to me.
Of course, this sharing and reuse makes package management and
dependencies more difficult.
For example, I need MD5, for which it seems to be necessary to load
Cryptography in Squeak, a quite large framework that seems to be under development (many unit tests failed, MD5 works OK). You have chosen a shortcut in WebClient by implementing MD5 yourself. This runs against the refactoring rule but makes it easier for users because they don't need to
load
a dependency.
Hi Sven:
Yes, Cryptography is mostly a name to group different implementations of crypto related stuff. I agree that may not be the better idea thinking in
reuse.
I'm copying this mail to the crypto list (even when I think that most of
the
people is here also) to discuss about better ways to name/install the
things.
Regards. Germán. _______________________________________________ Cryptography mailing list Cryptography@lists.squeakfoundation.org http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/cryptography