At 02:24 PM 9/10/1999 , Kester, Rush W. wrote:
>> > I am hoping someone can tell me I am wrong, and will point to some
>> > who supports a tool such as this (hosted on Linux, Solaris, or Windows
>> RRS had considered building such a tool commercially early on during the
>> development of Ada 9x (possibly as part of the development of what became
>> Claw). But the existence of c2ada made it likely that such a tool would
>> produce a lot of revenue, and we decided it we didn't need it for Claw. So
>> we didn't do it.
>> The thin bindings created by something like C2Ada don't buy you much
>> they're at least as hard to use as the C code, impossible to debug
>> (especially when something caused a mismatch of parameters), and don't do
>> much to show off the advantages of programming in Ada. Programming with
>> is like writing C in Ada syntax. To make them usable, you have to wrap
>> in a real Ada package - but then you might as well have built the whole
>> thing by hand. Which is why we built Claw...
>I disagree with those who say a thin binding doesn't buy you much. It let's
>you call a wide variety of third party libraries from Ada for which the
>library vendor only supplies a C binding. While it may not get you all the
>benefits of a strongly typed thick Ada binding, a C header to Ada binding
>generator would open a vast repository of third party libraries to Ada
In my case we have a number of projects that have created many programs and
program architectures written in C. Typically the Ada to C interface would
be very localized, but creating the record types corresponding to the C
structures is a big pain.
The best thing for users would be to have one tool licensed by all the Ada
vendors. This would make sure that the tool was available no matter what
platform or compiler was being used for a given project. This might be a
bit wasteful for projects that use multiple compilers (host and cross
compilers, for example), but presumably the cost for this tool would not be
terribly high, especially if it were sold on a per-seat basis. Of course
it would be even better if it were free, but I understand that programmers
need to eat, and therefore need to get paid.
In any event, I have my answer to my original questions (c2ada is the only
thing out there that anyone wants to talk about; it can be made to work to
some extent; no one is supporting it).
Draper Laboratory, MS 31
555 Technology Sq.
Cambridge, MA 02139, USA