directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Lecharny <elecha...@iktek.com>
Subject Re: Consolidate snickers projects into runtime and compiler
Date Mon, 21 Feb 2005 16:25:33 GMT
> > I'm confused, I do not see any work being done on the organization of 
> > ASN1 sub-project.  If there is a ADS standardization effort underway, 
> > then I'm happy to follow that.
> 
> I've been doing work slowly but surely towards this end.  I created the 
> rewrite branch to start pulling stuff from your branch into it and 
> bringing in some of Emmanuel's ideas at the same time.  I'd like to see 
> us all working on this new branch which we use to replace the trunk.

Hi guys,

I'm actually working on this branch (rewrite). I think that we will have
a kind of three layers stuff at the end: 
- stubs (generated by the compiler)
- digester or something like that
- BER/DER codec

Digester is the main point here. As Alex said in his wiki page, it
should be rewritten. Not that simple, because it carries all the
semantic of the LDAP grammar (and of others grammars).

The deeper you go, the more abstract you become. BER/DER codec should be
very context-free (no semantic at all, or at least, the semantic should
be carried by callbacks).

We have then two options here :
1) decode a TLV, even if it's a bad one
2) decode a TLV and check it immediatly (this is the way it's done
actually).

I think that the second option *must* be implemented to avoid a long
decoding of a twisted TLV.(a "fail fast" decoder).

The avantage of separating thoses three layers is that you just have to
agree on interfaces between them, and it gives you an opportinuty to
improve each layer separatly. Plus it allows to keep the whole thing
simple to deal with !

Pretty sure that Alan has a good work done on the compiler part.

Maybe we have to synchronize ourselves ?




Mime
View raw message