"Also we agreed that both groups would kep (sic) an eye on each
others proposals to make sure that the extensions did not
come into gross conflict. The full text of the agreement
is on the SID website (http://www.eda.org/sid) but I have
appended it to the end of this email anyway."
My point is that having such "side-agreements" is nice and I have
no problem with the division of responsibility, but it is
insufficient. The PAR language needs to address the issue
directly. The side agreements have no weight with the IEEE and
do not govern the scope/purpose of the proposed standards.
If the intent is compatibility, then what is wrong with codifying it?
Kamal also states:
"On the general problem: Of course, VHDL has to be seen to
be ONE language and not a set of extensions to a core."
and
"In the longer term, both these (SID and OOVHDL) will
be consolidated into VHDL but in the short term we should
let them grow properly before harvesting them.
And the harvester would be the VASG, right ?"
I disagree. VHDL is one language. It is IEEE 1076. VHDL-AMS is
another language derived from (or built upon) VHDL, it is 1076.1.
OOVHDL is not VHDL. It will be derived from VHDL. SID is not
VHDL. It will be derived from VHDL. (Note, I include within language
derivation the possibility that a new language is nearly 100% backward
compatible with its parent language. Just as C++ is backward compatible
with most C code.)
The analogy here is that C++ is not C, but is derived from C. C did not go
away when C++ became a standard. For the same reasons, VHDL did not go away
when 1076.1 became a standard and it will not go away if and when OOVHDL and
SID become standard(s). There is a large user base and a large base of
existing VHDL models. Many of these users will see no need to use SID or
OOVHDL for a significant period of time (perhaps forever).
I see no need for the VASG to harvest anything at this point in time. (Of
course, my opinion is subject to change given reality, but I expect the
timeframe to be greater than 5 years.) I anticipate that VHDL will grow old
looking very much the way it does now for the same reason C is growing old
without becoming C++.
This entire issue is at the heart of the OOVHDL and SID PARs and the scope
of VHDL 200x. The issue needs to be resolved ASAP.
Kamal also states:
"... this should not be confused with the mechanics of devising
new extensions. One cannot simply devise and validate
extensions to a language in large chunks. Smaller extensions
have to be tried out, often in parallel, and validated
before they are consolidated into the next version of the
language. And this is how I regard VHDL "dialects" - those
extensions of proven worth will get folded into the language."
If SID has not already been validated via ICL's VHDL+ work and OO has not
already been validated via the Objective VHDL work of Wolfgang Nebel, et al
and the SUAVE work of Peter Ashenden, et al, then why are we doing the
prototyping as a standardization activity? This is an inappropriate use of
the standards forum.
Of course, I'm being rhetorical here because (I think) both the SID and
OOVHDL groups would insist that the prototype work has been done and the
concepts proven. Therefore, both are at an appropriate stage for
standardization.
My points remain:
1. VHDL is a mature language with a large user base.
2. Therefore, VHDL cannot change significantly without significant negative
impact to that user base.
3. Both OOVHDL and SID are essentially defining the next generation
(successor to) VHDL (not the next version of VHDL).
4. To increase the chances of success for the next generation VHDL, the two
groups should merge their standardization and language design activities to
generate one language.
5. Failure to do so will risk the fragmentation of an emerging market into
segments too small to support tool and infrastructure development. (IMHO,
it guarantees that the market will be so fragmented.)
6. Failure to resolve this issue ASAP risks being fatally late to market
with a solution/product. Witness all the non-VHDL derivatives at this
year's DAC (C++, Java, Verilog/C -- also known as Superlog). These are the
competitors, not 1076! Can we afford to wait another 2-5 years for separate
standardization of OOVHDL and SID and then another 5 years after that for
the two to be merged? By then, VHDL will be an old language with a
shrinking user base and OOVHDL and SID will have been relegated to history's
ash heap. Of course, merging the two activities does not guarantee success
of VHDL TNG (VHDL The Next Generation), but failing to merge does guarantee
failure (IMHO).
-Stephen Bailey
Chair, IEEE 1076, 1076a