I'm not a developer (but I have aspirations :), so the things I say on this issue should not be given too much weight, but here's my 2 cents nevertheless. If I say something stupid or uninformed, please correct me so I know better.

It is possible, but I don't prefer to do so due to the lack of developers.

Totally agree. At this point, with the number of developers we have, specialization and not duplicating each others' efforts is extremely important.

  Relying on a single, mature library for each purpose is better in my
point of view.

Some criteria I can think of for how we choose libraries:
- Maturity -- the more mature it is, the fewer likely bugs there will be - Popularity -- the more popular it is, the larger the support network, and the more likely new developers will have already been exposed to it (lower barrier to entry) - Development Language -- How much work will we have to do to support the library? The way we rate a library could be as follows, with the most favor given to those higher on the list
      a) Third-party developed natively in Objective-C
      b) C (we would maintain ObjC wrapper)
c) Rewrite from other language (Java, C#, C++) where we need to maintain all

  We just need to choose one.
  For example:

  Built-in Database: db4, SQLite/SQLDatabase or QDBM ?
  Regular expression: regex or OniGuruma/OgreKit ?
  Compression: zlib or other choice ?

Definitely zlib, especially if we're interested in using the OpenDocument (OOo) format as the base document format (I suggested that previously). The OpenDocument formats use zlib to compress a folder into a single file/"bundle".

  Indexing/Searching: lucene/LuceneKit or QDBM/Odeum ?

I would say Lucene. Even though QDBM might be easier to maintain, Lucene is having great success in the industry right now and using it might give Etoile more to leverage for marketing. Of course, I have really no idea of the technical advantages, but that's my opinion as an "advertising-guy". Maybe not the best basis for a technology choice...

Flame on!


J.





Reply via email to