Mickael,
Comment below.
On 22.11.2016 21:26, Mickael Istria
wrote:
On 11/22/2016 08:33 PM, Ed Merks
wrote:
We considered it, yes. Then we considered what it would be
like to spend many months on politics, backwards compatibility
testing, arguing about design points, and so on, and so on,
verses simply spending several weeks implementing something
better that can coexist will without adverse impact on legacy
details. Decisions, decisions..
So if I understand correctly, contributing to PDE would have been
the ideal case technically speaking, but happened to be too
expensive, even for contributors as experienced as you and Eike?
No, technically speaking it doesn't much matter where such
technology lives. We did not have a need to change PDE itself.
It's extensible enough as it is and the legacy implementation is
inflexible and brittle and not amenable to significant improvements.
I can for sure understand the argument why not
contributing to PDE first in that case, and I hope the cost of
contributing to an established project such as PDE gets lower.
In the end, the serialization of the .target file is
a direct reflection of the poor implementation (model) of the
ITargetDefinition. You can't simply improve the editor nor can
you implement a good design from the existing serialization.
It's just legacy, and in the end, the old design needs to remain
there for compatibility.
There are strategies used in Platform to extend model/APIs without
breaking existing stuff. It's not an easy problem, but it has
solutions (which aren't easy neither).
Exactly, I know those solutions well. There are a lot of I*[23456]
names involved. And what's the point in the end? A different
implementation and a more flexible model solves the problem better
than the layers of baggage-approach. That was exactly your point
about one layer of abstract too many, but it's clear that such
layers don't matter as long as the whole triple-decker sandwich is
in one project.
Unfortunately, without really understanding it, some
feel free, to characterize, in a
most-humble-of-opinions-sort-of-way-of-course, as one layer of
abstraction too many.
Did you consider looking more closely at all the details
before expressing a most-humble-opinion
There is no need to go into details to understand that there is a
strong duplication of concerns between all those TP formats; and
that multiplying solutions for the same issue in a community such
as ours can become an anti-pattern as it spread resources across
multiple projects and reduce consistency and can even increase the
integration cost (if Tycho wants to support all those format, it
multiplies the effort, whereas if all were in PDE, there would
probably be an API able to consume any of those and integration
would have to be done only once for all those formats.
Tycho would only need to use the API, but that's not the design
approach it has taken (I believe), so it too is an example of
multiplying solutions.
In the beginning of your answer, you explain why you didn't
contribute it to PDE; and there was no big technical blocker, it
was all about need to find levels of agreement with current
project state and "time to market". So somehow, it seems like
yourself do agree that the TP work in Oomph would make sense to be
in PDE, and that at that point of time, it was/is not something
you were willing to do there - for the sake of productivity.
Not really no, not unless Oomph pretty much as a whole is moved into
the platform, or lower still. E.g., there is
https://bugs.eclipse.org/bugs/show_bug.cgi?id=500329 to suggest the
nice repository explorer could/should be moved to PDE, but it
doesn't depend on PDE at all. It's purely p2 technology, so it
could be moved to Equinox. But I notice that no one on the platform
team is keen on that suggestion. Go figure huh?
Overall the question is: where as a whole does the Oomph technology
stack live? Why does it need to move at all?
So at least, I read your mail as an acknowledgment
that the TP parts of Oomph would rather be in PDE, so we somehow
agree.
No, you read that wrong.
Maybe it would be the right time to consider
migrating some pieces of Oomph to PDE?
No, read 500329 and consider where the Oomph technology stack
belongs.
_______________________________________________
eclipse.org-architecture-council mailing list
eclipse.org-architecture-council@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse.org-architecture-council
IMPORTANT: Membership in this list is generated by processes internal to the Eclipse Foundation. To be permanently removed from this list, you must contact emo@xxxxxxxxxxx to request removal.
|