| Michael Erdmann 2004-09-28, 4:13 pm |
| Anton van Straaten wrote:
> Michael Erdmann wrote:
>
>
>
> I think that'll depend a lot on who the developers are. On some of the
> Scheme projects I've seen or been involved in, those kind of choices get
> made by the team members.
May be the problem is that i am still following a wrong paradigm, which
means you look at your requirements, you break them down in managable
parts, define interfaces between them, Manageable part mean that the
same developer can work on theses components incrementally or several
developers can colaborte where each developer has been assigned to
each part. In the past i found that some what stable interfaces between
these parts are an essential issue. Stable means not frozen since
you can never avoid that interfaces are changing. Stable means, if
something i changed everybody in the team has a chance the realize
the change because the compiler is complaining. I am realy afraid
that a not trivial project is simply falling aprt during the time,
and every thing has to be tested.
I am realy wondering what choices i have. In the Moment i got the
following issues.
1. selection of the correct scheme environment.
bigloo, chicken, SIOD, DrScheme ...
2. Tools for impact analysis, when i am for example changing the input
argument of a function from a single value to a vector (structure).
Since the compiler/interpreter is not warning me about any in-
compatablitiy, i guess i have to rely on fgrep, an editor and awk?
Maybe SoftScheme or what else?
3. Documentation.
This is in all languages a problem. But what it the most commonly
used tool in the scheme community. I don't like to reinvent
this tooling.
Well i am trying to get schmozz with chicken running and i am
failing misserably.
4. Build environment
Are you using the GNU config tools. For example the makefile of
SLIB does not use config. What is the common habit here?
5. Deployment
What is the typical way of deploying. Are there some expriences
and methods available (rpm, tar achrive what so ever). Specially
the issue that you need to deploy the interpreter or runtime
system as well!
6. Coding guideline, which means in any language i can do everthing
(e.g. c-code like (++p* = *( get_address(*q++) )..) but there
resonable limitation specific for certain languages.
7. Typical design blueprints like client/server, data base application
etc. Any thing i should never do, since one million developers
already have failed in trying it.
8. What is about embedding into operating systems and grafical
environments (GtK etc..)?
9. Portability.
But i think this only a minor issue, if Issue 8 is solved on
different platforms.
> If you're talking about a single maintainer
> initially and hoping that you can get someone else to take over in future, I
> don't think the exact details of your Scheme choices will be the real issue.
I am not so sure. If some body takes over and i have reinvented
everthing from the scratch it will be ver difficult for him to
get even the build environment running on his box. I had this experience
in the past very often.
> Whether the project itself achieves something worthwhile is more important.
> After all, people work with the Scheme implementation SIOD as part of the
> GIMP, for example, and SIOD is far from being one of the best or richest
> Scheme implementations. More on this below.
May be this an answer for Question 1. I will check out.
>
>
>
>
> Google shows a copy of SoftScheme here:
> http://www.bloodandcoffee.net/campb...soft_scm.tar.gz
Thanks, i have done the same, but obiously missed it, sorry!
>
> As for projects implemented in Scheme, the Roadsend PHP compiler was
> mentioned, implemented with Bigloo Scheme, but I don't think that's an open
> source project: http://www.roadsend.com/
>
> Randomly searching Google, here's an example of Scheme being used as an
> interpreter for Open Inventor: http://xenia.media.mit.edu/~kbrussel/Ivy/ .
> This project seems to have a similar overall technical structure to what
> you're trying to do.
Thanks, the Header2Scheme tool might be interesting. I will have
a look in this.
>
> There are some Scheme libraries for PLT Scheme maintained by the Schematics
> project at http://schematics.sourceforge.net/ .
>
> As I think someone else mentioned, you might also look at some of the Scheme
> implementations themselves, since they tend to be large projects involving a
> lot of Scheme code.
Yes, this was the reason why i wanted to get schmozz working with
scheme. To me scheme look far to complex for me, but let see how
i am doing.
>
>
> Different Scheme implementations have different tools that are commonly used
> with that implementation. This applies to documentation tools, module
> systems, even static type checking. If you pick a Scheme implementation
> with an active community, you might check which documentation tools are used
> most commonly in that community. For example, I imagine you're more likely
> to find Schmooz being used with SCM, and PLT has its own system for
> generating documentation for its Help Desk (and the web). I see that
> Chicken has a package for the documentation tool "mole".
May be this is what is drving me crazy. There are so many scheme
implementation. In my simple GNU world everbody is using gcc, gnat
and some GNU tools. In the scheme communitiy there seems to me an
artifical amount to implementations and tools. Only a view of them
are regulary maintained.
Thank you very mutch. I guess what i will do by end of this w . I will
try to build a small protoype of refactoring tools i am intending
to build using chicken assuming that i can incorperate the dtcall
interface into a chicken extsion. During this i will try to test most
of the techniques i find important for coding and deployment.
During the w end i will initiate a new thread putting my results
on the 9 topics aboe for discussion. Next w i will then decide
wheterh i am going to proceed with scheme or if i am connecting
framerd to C# (mean mono).
Michael
>
> Anton
>
>
|