[Chandler-dev] [Sum] The Great Architecture Discussion of 2007
vajda at osafoundation.org
Tue Oct 9 22:39:34 PDT 2007
On Tue, 9 Oct 2007, Phillip J. Eby wrote:
> The thing is, once you look at this from the app layering perspective, the
> mismatch between the relatively simple things the app is trying to do, and
> the very powerful generality of the repository, becomes more apparent.
So, I ask again to the general readership of this list - Phillip, I think
your opinion is pretty clear - do we want to compile Chandler into a
"relatively simple" app like Mail.app or do we want to keep the "powerful
generality" part of the vision and implementation of Chandler when I joined
the project in 2003 ?
>> If that's the route we'd like to take Chandler to, fine. That should be
>> clearly stated.
> I believe Katie has already stated it, even as far back as the creation of
> the schema API. But I imagine she will clarify it again if needed.
Katie, I certainly missed your earlier statement in this direction.
What is your take on this ? Where do you stand on this ?
Phillip's opinions in this area are not news to me. I even offered him my
repository job shortly after he joined OSAF when we had a similar
conversations. For various reasons, he didn't take it then. I'm ready to
make the same offer again as we're ready - it seems - to abandon another -
major this time - part of the early vision.
> I mean that by implementing a skiplist *inside* of BerkeleyDB rather than
> using a native BerkeleyDB structure, we're adding an "interpretation" layer
What is the native Berkeley DB structure that corresponds to a skiplist ?
A skiplist is just a fatter ref collection (fatter because of longer
reaching links between nodes). Are ref collections also duplicating an
existing Berkeley DB structure ?
> Yes. The key distinction vis-a-vis relational vs. repository, is that we
> would now be adding another Python layer, in addition to the those that
> already exist. Whereas, if we used a Python ORM, we would have just the
> mapping and an all-C backend that we don't have to maintain. In fact, the
> mapping layer might also be maintained by someone else, if we use one of the
> many O-R mappers for Python such as SQLObject, Storm, Axiom, DejaVu,
> SQLAlchemy, Mother,... and probably others I've forgotten about.
The repository back-end is already mostly C. Many of the places where the
repository code was spending time were moved to C. There is more C code than
python code in the repository by now. Most of the time today is spent in
firing thousands of useless notifications. Loading and committing items is
pretty fast in comparison. (Merging items is still pretty slow, though)
I don't know where you're getting that the repository is slow in doing its
persistence job. It slows down in its mapping layer, in the layer between
the persistence layer and the app. We have so many sorted indexes because of
the generality of the Sidebar that whenever a trivial change is made in an
attribute, they all have to be indexed. I don't see where this work comes
for free, either code-wise, or perf-wise, in a relational model. If you
cache the results of sorted queries then you have to maintain these caches
there as well.
By killing Chandler's use of the existing powerful/general repository you can
also kill some of Chandler's powerful/general features, thereby fixing the
perceived performance problems that plague it today. This is certainly a
way to solve the problem.
OSAF management and developers, is this what we want ?
More information about the chandler-dev