Lambda the Ultimate

See this comment, though I doubt #ltu gets much traffic since I've not seen it mentioned since (if anyone uses it, please correct me).

I have no opinion about this one way or the other.
Keep in mind though that there is a close correlation between the facts that "detailed discussions regarding design and implementation of programming languages and compilers are considered "out of scope" in the policy document" and "LtU attracts a lot of really bright people".

The policy wasn't something that was mandated when LtU was born - it was the result of those bright people complaining, and asking for policies restricting unsubstantiated discussions.

It seems that many people interested in programming languages come across LtU and discern value, because of multiple qualities that they don't find anywhere else (member expertise and prominence, site participation and liveliness, and the fact that "programming languages" and "theory" are mentioned in the same sentence). The sad thing is, the "average programmer" doesn't understand the difference between computer science and programming languages, so they see this as the place to discuss anything and everything programming-language-related, whether or not the theoretical notions are strictly textbook computer science material. It seems that the phrase "programming language theory" blurrs the distinction between programming-language-related and computer-science-related subjects. LtU could do more to make a distinction, probably by banning the use of "PLT" and "programming language theory", avoiding the use of the terms "programming" and "programming languages" as much as possible, and by emphasizing "computer science", "lambda calculus", and other such terminology. Otherwise, attracting programmers who are writing or thinking about programming languages (as distinct from those knowledgeable in classic computer science topics), just comes with the territory. It's a trade-off between raising general awareness of the field of computer science, and being strictly on-topic but "nichy and obscure" (from the perspective of the "average programmer").


An afterthought: perhaps changing the sub-title from "the programming languages weblog" to "the computer science weblog", would do alot to send people a quick message about the subject of the forum.


Of course, like many things, it all depends on your goals & roles for LtU. (I wonder if the policy statement would be better articluated in terms of goals & roles for LtU? An interesting idea to explore...)


One thing to remember, is that LtU is (in experience) a sort of melting pot for a few different camps: programmers, computer scientists, et. al. It's audience is broader and larger in that way. There is a value to this, in that there is a cross-fertilization of knowledge, and a proliferation of awareness, that occurs across the field(s).


But, a separate weblog for highly prominent and experienced programmers actually might be good for the world, or maybe LtU just needs to have 2 broad "arch-categories", which can be used to optionally narrow results while browsing, thus still providing outlet to both programming languages and computer science at LtU, to those times we want both.


I wrote more than I thought I would; glean what you will.

No other comment.

I'd very much like to see this. It's trivial to set up another forum, and even more trivial to kill it if it doesn't work out.

First, it seems to me that some implementation discussions are aleady welcome here. Witness past discussions of CPS vs. ANF intermediate languages, my own recent question on efficient implementation of structural types and so on and so forth. I'll surmise that when implementation issue bumps up against an interesting PLT topic, folks will gladly discuss implementation here.

OTOH, second, it would be more "clean" to segregate the PLT related implementation discussions into another forum where folks can be a bit more free wheeling. In my own implementation oriented inquiries, while appreciating the generous responses, I've been reluctant to delve more deeply, ask follow up questions, etc.

For my part, I have many questions likely best answered here in the LTU community. For example, there are complicated issues of implementing unboxed parametric types I'd love to discuss, the related issues with "perfect" GC, type erasure, BIBOP related options and so on. While "unboxed parametric types" might be a reasonable LTU topic, the meaty follow on discussions unfortunately might not - which is a shame.

So I vote for a "Putting PLT to Work" oriented sub-forum. Nothing to lose and possibly lots and lots to gain.

Scott

I'd also welcome informed discussion related to optimization and code-generation that goes beyond current practice. It seem natural to allow themes as those in a "Putting PLT to Work" sub-forum or sister site.

I think I should emphasize that implementation techniques and design discussions are two radically different things. Implementation issues were always on topic for LtU - indeed, there is an "Implementation" department, as well as one of "cross language runtimes" (surely an implementation issue). These topics are visited rather less frequently than others - mainly, I expect, because most contributing editors find them less interesting than other topics (I belong to this camp myself). However, if something new and interesting comes along it will surely be discussed.

Design discussions, however, are something else. I think one of the reasons LtU attracts people with different backgrounds and interests and remains a productive forum for discussion, is that most discussions are anchored to specific scientific papers or project, which ground the discussion and ensure some level of scientific rigor. Design discussions that did take place here tended not to rise to the level of discussion many here want. They often begin with underspecified questions, various assumptions that the poster finds "obviously true" come up as aggressive arguments deep into the discussion, the goals of the discussion are unclear, people have vested interests in their own projects etc.

My feeling is that if someone feels he can post links to interesting work on implementation techniques, they should sign up as contributing editors. LtU is shaped by the community, and we appreciate the input from all contributors. On the other hand, I don't think LtU is the right place for design discussions, which are better served by project specific sites and mailing lists. Posting a short message here inviting people to join such discussions that occur elsewhere is of course welcome.

Anyone of course can start a site for design discussions. I don't think it will attract most of the LtU readership, but if it does it would be a sign that the interest is real, and it would be great to have such a forum for discussion (though, see links to sites and lists that already attempt to fill that niche).
That's at least how I see the situation. More comments from LtU regulars are of course welcome.