[9fans] programming languages

[9fans] programming languages

[forsyth]

>>operators were useful but not as important.  I would have
>>been happier with Limbo as a first class language on
>>Plan 9 than something that runs under emu.  A more
>>integrated environment on Plan 9 would be cool.

Indeed.  The thought had crossed our minds.  More precisely, as I see it, emu
has two main functions: to provide the namespace
and device interfaces that are accessed through Sys,
and to provide the interpreting/gc/module support
for Limbo and other languages via Dis.  It's both an operating
system (as application) and a run-time support for Dis/Limbo.
Only the latter should really be needed (in a logical sense) under Plan 9.

The other aspect of the Inferno environment is the
content and layout of /usr/inferno.  If the names chosen had
been identical to those of Plan 9, there wouldn't be
any need for that either (on Plan 9).

Having smooth integration with Plan 9,
whilst retaining the reasonably high degree of portability
that Inferno provides for Limbo applications across all
platforms (even though there are flaws) will require
care, but should be possible.   (The portability really is important
for cross-architecture, cross-system distributed applications, which
in turn is one of the things Inferno can do well -- Erlang is the
main competitor there, but doesn't cover the embedded space,
and Java is just generally hopeless.)

>>Other features like strings as a firt class type with their own
>>operators were useful but not as important.  I would have

I find that, and garbage collection, to be quite important in
making it easier to write concurrent applications without strain;
the reduction in programming effort is not as great as the introduction
of static type checking but it feels much the same to me.
``Who is responsible for freeing this value?''
Of course, you can provide your own reference counting
but then again, you could also provide your own type checking (at run-time).

Overall, we've developed some computer science, especially in
type systems and concurrent programming, since C (or even C++)
appeared.  Limbo at least nods at it.  Java's approach
to concurrency is pathetically stuck in a 25-year-old time warp,
and worse, they never untangled the clash between concurrency
and object-orientation (unsurprisingly, because at least until recently
it was still a research topic).

Re: [9fans] programming languages

[Boyd Roberts]

> Java's approach
> to concurrency is pathetically stuck in a 25-year-old time warp,
> and worse, they never untangled the clash between concurrency
> and object-orientation (unsurprisingly, because at least until
> recently
> it was still a research topic).

yeah, i saw a doc on java and they were interviewing the 'designers'
and one of them even admitted that java was never designed to do
what it is now used for.  that about says it all.