🚀 go-pugleaf

For my master's thesis I am planning on investigating techniques for the
implementation of Directory Information Bases.  As a bench-mark I would like
to compare my work against QUIPU's DIB implementation for a couple of reasons:

1) QUIPU is probably the most well-known implementation.

2) If possible, I would like to try to improve the performance of QUIPU and
   contribute the my results back to the ISODE community.

If people have an interest in this work and/or can provide me with information
and references to related work, I would greatly appreciate it if you could
contact me by e-mail.

I have already done some investigation on implementing an OSI Directory Information
Base using existing database tools: ObjectStore from Objective Designs Inc. and
Sybase (both commercial products).  My main thesis is that it should be possible
to do at least as well or better using existing database systems to implement a
DIB.  Also there are additional advantages to using existing DB tools than a
custom DIB implementation.

Some of the goals include:

a) Reduced start-up time.  I'm quite confident I can eliminate the long start-up
   time for QUIPU to load all the EDB files.

b) Reduced update/insertion time.  I'm also confident I can reduce the time it
   takes to add or change information in the DIB.

c) Improved access time.  This is the really interesting area--since both database
   systems are disk-based, my goal is to do at least as well as, preferably better
   than, QUIPU's memory resident DIB.  Since both Sybase and ObjectStore support
   concureny in queries, I'm hoping that average responce time can be improved.
   Does anyone know if QUIPU is single threaded or multithreaded?

d) Time permitting, explore some of the distributed schema aspects of X.500/1992
   in implementing the DIB.

Ideally I would like to design a couple of new DIBs and `plug them into' QUIPU.
I'm not sure how feasible this is in the time frame I have.  I've been looking
at the code of ISODE-8.0 trying to understand how QUIPU works, but...  If anyone
could answer some questions on QUIPU I'd really appreciate it.

1) Why are there no comments in the source code?  Is the code documented somewhere
   else, were the comments stripped out for `intelectual property rights' reasons,
   or is that just the way the authors wrote the code (no insult intended)?

2) Is there a fairly clear division between the QUIPU network/protocol engine and
   the DIB implementation, or is everything rather intertwingled?

Eric Kolotyluk (eric@sfu.ca)
School of Computing Science
Simon Fraser University