Re: Continuation based app servers

my take : (this is in no way to criticize ucw, just something that i
observed)

1. steep learning curve

so you thought you finally get a handle of CL, and now you're
presented with a whole new set of things to learn call/cc / cps
transformation, etc

2. small community

you might not be able to get help quickly.

but the major problem is that since there are not many people using
it, there's no pressure to make it run across all major
implementations (acl / lispworks / cmucl / sbcl / openmcl / mcl / ecl)
x os combinations (linux / windows / macosx). this is often the most
frustrating thing that can put users off (it's a common problem across
most of the cl libraries, not just ucw)

3. documentation is thin

for most laymen (that's me) they can follow the example codes, tweak
it a bit, but usually they can't get far.

4. leaky abstraction (http://www.joelonsoftware.com/articles/
LeakyAbstractions.html)

well, at least with scheme, you know what a continuation is (but it
already introduce a lot of problems that you don't have otherwise).
with ucw, you need to understand how its kall/cc different from
scheme's call/cc and when and where you can't use it.

the worse thing with leaky abstraction is that, if the app works fine,
it's ok. but if it drops to the debugger, you better understand ucw
_really_ well.

that said, it didn't stop very smart people like drewc (who creates
LOL) and folks from para-agent (who made a killer app w/ ucw) to build
something really complex with it.

i guess ucw is like cl, if it falls on the hands of an expert, it can
be really powerful.

my 0.02


.