directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Lecharny <>
Subject RE: [asn1] These ideas are exceptional Emmanuel!
Date Sun, 20 Feb 2005 00:01:08 GMT
Le vendredi 18 février 2005 à 16:08 -0500, Noel J. Bergman a écrit :
> > A while back we had an IRC discussion on why we needed both general
> > TLV processing and compiled smart stubs.  Does anyone recall what
> > the argument for that was?
> I checked my logs, and have some commentary on TLVs and POJOs from mid-July.
> Nothing more recent.  Would that it be it?
> 	--- Noel

It may be two weeks ago, when IRC wasn't logged? We had a discussion
about "smart stub" that are pre-compiled TLV which can be sent without
encoding, as they don't change. For instance, a sucessfull BindResponse
will always have the same bytes, except for the MessageID. So we can
pre-build the TLV, and just push the current MessageID in it. It works
for Response, not for request (it could, but it's useless, as
performance issues will occur on the server, not on the client).

Of course, depending of the MessageID, we will need more than one of
these "smart stubs" - TLV Patterns is a better name for it -. You can
then push a buffer without having to encode a TLV. 

That was the idea, if this is the conversation you had in mind.

btw, Alex suggested that we could build two kind of TagDecoder, one for
single byte tags, and another for multi-bytes tags. I think that it's
not necessary, as the single byte tag Decoder will always run first,
followed by the multi-byte Decodcer, depending on a specific condition :
b[0] & 0x1F == 0x1F. This test is absolutly mandatory, so even if we
don't have a multi-byte tag, we won't be able to avoid this test. 

View raw message