Lambda the Ultimate

I wasn't too sure if this was the best place to come with a question like this, but I don't have many (ok any) contacts in the Lisp (CL or Scheme) world that are as well informed and opinionated as the readers of this site.

I have read many times that the programming environments of the Symbolics machines were some of the best environments to work in when programming in Lisp. To my surprise, I haven't seen any modern environments attempt (or currently attempting) to reach this level of productivity. Or if they are attempting there is a magic missing from them. But more importantly I haven't seen any commercial company attempting the development of such an editor.

Now mind you my experience with these environments is less than 'nil considering the fact that the only reference I have to go off of are the Rainer videos. I feel like a kid trying to experience the true magic of the Wild West by only watching John Wayne movies. But I can deal with the lack of experience, but a commercial project cannot not be done without the voice of experience and/or customers. This is where some help you the readers of this site can possibly help me, if willing.

I have two problems, one is the runtime. My first instinct jumps to mzscheme, because of its rather large install base across universities. I truly enjoy programming in Scheme, and see many competitors in the CL market that I am not to sure I could contend with. So my first question, with consideration from a strictly pragmatic viewpoint, is which environment is better for the user, or rather which language is better suited for what job. I know people have their preferences, but I guess I would like to know why they lean one way or the other. Or if both are used, which language is used for which types of jobs.

The second question concerns the graphical front end (more appropriately the 'product' piggy-backing on the language). I was enchanted with the graphical environment of the Symbolics machines, and just recently discovered that this set of widgets was a standard called CLIM. I also recently found out that McCLIM is a nice project that is working to create a LLGPL version of the standard.

As you can imagine finding this only put more decisions into my head. The choice was easy when I was only dealing with MrEd. I am looking for recommendations as to which road to travel, or rather to present some options I may not be seeing. On one side I feel that I could use MrEd, and strictly develop the entire environment in Scheme using its libraries. I am sure that I could use the FFI to call CL environments if the customers wished us to offer a CL version of the editor. My concern (due to lack of experience with MrEd) is that I won't be able to get the same 'Symbolics magic' from MrEd that may be possible with CLIM.
On the other hand, it is also possible to develop the editor using CLIM and CL, and use the CL's FFI implementation to call mzscheme to evaluate Scheme code. Both are viable solutions, but I see a bigger problem.

The Lisp family of languages grants much power to its user on a metalevel. This was the main reason for me wanting such an environment; no IDE developer has been able to guess every need I have (hell I don't even know until I know) so I wanted a programmable environment like those of the past (the ones everyone seems to rave about). The bigger problem I mentioned above, when using the combined language is approach, is that potential loss of programmability. If the editor is in CL and I am only offering a Scheme product, then its possible that the editor wouldn't be programmable (without a CL environment thrown in as well) which defeats the purpose of working on such an editor.

I do apologize for the length of this posting and if the content is not suited towards this site. Thank you for taking the time to read it, and I appreciate any constructive comments or advice you may have.

Best Regards,

MJ Stahl

I have recently found scheme to be very expressive language that allows me to not only quickly implement my project, but also to extend it easily. I also need my application to be as fast as possible (its basically a database application, as mentioned in an earlier thread). From what I gather, scheme/lisp implementation are far slower than C or C++ (even OCaml seems very fast). I've seen benchmarks where even scheme-to-c ends up being several times slower than c++.

Now my question is this: is it at all possible to bring the performance of scheme closer to c++?

I would think that lisp/scheme would actually be easy to optimize. Could they be optimized further if haskell like typing was introduced (I think Qi does that)...more information available to the compiler, better it can work!

If there are some things intrinsic to lisp/scheme (perhaps eval?) which keeps performance low, then I'll have to live with some performnace upper limit. Otherwise, why don't we see faster implementations? Is it because those who use lisp/scheme just don't have a need for performance?

I don't even mind if current performance limits exist simply because no one has written an aggressive enough compiler. I can prototype my application...then start work on a compiler.

As you can see being quite busy I have to reduce my contributions to LtU. I guess I am not alone in this regard as the frequency of new items mentioned on the home page shows.

I urge editors to post more often. Several items posted to the LtU forum were quite relevant for the home page, especially due to the low volume of new messages. Remember that if you feel an item in the forum is worthy of wider attention, you can post a link to it on the home page.

That concludes the message intended for current contributing editors.

If, howver, you are not an editor, yet visit LtU regularly and wish to become one, get in touch with me, and get added to our illustrious team.