Hello folks,

just a word to say that I agree 100% with all what @Joel said, and IMHO 
Camp2Camp is one of the very best real OpenERP partner reference as they jumped 
in in early 2006 (you crazy folks), survived it and were even pretty successful 
now.

I insist: shorter release cycles really doesn't mean more bug, because as Joel 
stated it's much easier to delay risky changes all together to the next coming 
release rather than cheating on the stable branch as you are forced to by 
customer needs.

But since next version is coming a soon enough, more partner can actually test 
it/fix it BEFORE it's released, so those risky changes are actually tested and 
this would IMHO rather meam LESS bugs. This also mean more people would put 
resource developing OpenERP, it actually mean less internal costs for Tiny or a 
faster progression instead.

In my experience, it would rather cost less to do two small migrations (even 
more if there are tests) rather than a huge one like 4.2 to 5. IMHO, the main 
reason why 4 to 5 scripts are still not available for the community is NOT a 
new profit strategy as some accused Tiny of. It's rather that changes were so 
huge that even Tiny was unable to release the kind of handy migration scripts 
they were releasing previously at the same costs (thus releasing them for free).

And what is the best: 2 small and free migration scripts lot's of people will 
test and improve or no script at all, and pay Tiny or some partner to see what 
happens like v4 to v5 migration? BTW, the issue is that customers will hardly 
be willing to pay the real price of the v4 to v5 migration cost because they 
can't imagine it's so complex, whereas they might pay (less) for two small and 
smooth migrations, or more probably, one yearly migration that would consist in 
two steps for the partner but would, I think, cost less.


@sraps: don't be that suspicious, there isn't any dark world behind partners: 
except training, I think there is no tech information at all partners have the 
community doesn't have (so yes it's tough but we do it).

I agree with you about the API thing: API changes need to be announced (almost 
OK even if API blog policy not followed, a resource issue I guess) but also 
planned and discussed. Again, by releasing more often, API changes could be 
concentrated all together (and explained/tested) at version changes, rather 
than on 'stable' version is it is the case now.

Moreover, currently, even if Tiny would like to delay all API changes they 
think about to the next yearly version it wouldn't be possible. Why? Because 
they would have only 2/3 months of heavy resources to put in building the new 
version (nobody outside helping as they have no incentive in doing so with long 
release cycles) so they would actually lack resources. And if they do a mistake 
it will last for one year. Finally they would even forget about the changes 
that were discussed one year back (I just imagine a huge list of un-classed 
blueprints with no comment/feedback), just like with the bugs, probably only 
the last ones would be implemented.

Finally I agree with sraps about the fact that unfortunately module API is not 
so well designed. It would need to change... How to improve it? If you release 
too slowly it would take lots of years and you might changes API as customer 
needs arise on stable branches. And for most of the API it would mean that 
useful refactoring would simply never be done.

What are the implications of not doing that API refactoring? Well, IMHO the 
implication are less module integration (less usability) and MORE bugs. Indeed, 
improper API design means that module implementers need to cheat the API (just 
like in for the product configuartor I made at Smile where I had to override 
ir.actions for instance...) , do things they shouldn't do, copy/paste large 
portions of code (they would lately received fixes unless in the pasted 
versions).

For instance, if a method is too large, doing lot's of things in the DB and not 
returning all the IDs it created/changed in DB, then if you override you should 
do hugly things to UNDO part of what has been done or copy/paste a 150 lines 
methods that will never receive fixes in your version... It's very important to 
carry on the refactoring over the time an plan it in order that it doesn't 
become a cost for all users.
Please read that blueprint for details:
https://blueprints.launchpad.net/openobject-addons/+spec/code-refactor-for-better-extensibility


All right, again v5 is really usable and good by now, it's just unfortunate it 
took so long to get it bugfree and we all hope next version won't take as long. 
If it's the case, just because v5 is good enough, then I think lot's of people 
will delay v5.2/v6 adoption as long as possible in order to avoid being the 
first kamikazes, meaning that the one that would do the move would suffer even 
more. For instance, well advised Camp2Camp has been one of the last to move to 
v5. In one hand they were right cause they avoid crash testing for them, in the 
other hand we lacked their help. I would rather like we could all help in 
stabilizing the next versions.

Best regards

------------------------
Raphaël Valyi

CEO and OpenERP consultant at
http://www.akretion.com




-------------------- m2f --------------------

--
http://www.openobject.com/forum/viewtopic.php?p=42493#42493

-------------------- m2f --------------------


_______________________________________________
Tinyerp-users mailing list
http://tiny.be/mailman2/listinfo/tinyerp-users

Reply via email to