[jcifs] Re: Tridge's PIDL compiler and the Samba4 IDLs.

[Michael B Allen]

Christopher R. Hertel said:
> Why can't the PIDL back-end produce the stubs you need?

First, if you look at what PIDL generates and what idlc generates the idlc
stub is much simplier (albeit a little bigger because everything is
unrolled/inlined but that may be reduced in the future, NetShareEnum.java
is 185 lines w/ 2 info levels). The idlc code necessary to generate an
arbitrary stub from the parse tree is currently 273 lines of C. I think
there are some tricks to NDR (described in a previous message here) that
Andrew may not have handled quite right and I'm worried that that is the
source of these mysterous "special cases" that require "tuning".

Second, Jarapac/idlc is a generic toolchain. We are not just interested in
Windows management functions. I want people with MS-RPC and DCE projects
to be able integrate Java easily.  I want high-compatibility so that it
can be used for things we didn't anticipate. That's when the interesting
stuff happens. Otherwise why are we doing this? So we can list Unicode
share names. No. With a little luck, people will be able to basically copy
a stationary, run 'ant compile' on their interface definitions, and get a
jar and API documentation. Using PIDL would require converting all of
their interfaces to the PIDL syntax.

Mike