[Building Sakai] K1 - uitils rationalization?
csev
csev at umich.edu
Mon Jun 8 07:15:10 PDT 2009
Rationalization is a great idea and deprecating is OK - particularly
if the messages are really clear and perhaps someone writes a document
page about how *exactly* to switch to the new/right APIs.
I think that at some level we are getting *well * the point where non-
upwards compatible API changes to the 2.x / K1 code base have a
strongly negative cost benefit regardless of how "elegant" the changes
are.
With the Sakai 2/Sakai 3 already splitting our energies and many
schools assembling their production build from a combination of core
and contrib and their own repos - we need to make the lives of contrib
developers as comfortable as possible for the rest of the 2.x / K1.
We have seen the impact of the K1 refactor (something wonderfully
elegant and something I strongly advocated for in 2007) and how long
it has taken to release 2.6 - because the elegant changes were done
and about the time there was no going back - the primary K1 resources
all left onto other projects.
I think that it is time to accept the fact that the 2.x code base
needs to take a VERY conservative approach to non-upwards compatible
API changes.
At the same time, I think that Sakai 2.x has a long and happy life
ahead of it - and needs investment and improvement - but we can work
on things that make the software better/faster/more reliable/more
usable *without* improving the elegance of the APIs.
So my opinion is that refactors of APIs are perhaps our *lowest*
priority and perhaps the most harmful of all steps that we can take at
this point in the code lifecycle.
If my vote mattered I would vote -1 even on deprecating.
Why bother? What is*broken* that needs fixing? How many
conversations to explain the changes will have to happen about this
from our far-flung developers and folks who want to pull bits and
pieces from trunk for their production???
/Chuck
More information about the sakai-dev
mailing list