Dan wrote:
As far as using jacl, I couldn't disagree more. Tclblend is
designed to integrate with c-code, so it's a nice fit with aolserver
as an independent module. To use jacl, you would have to
reimplement aolserver in java, which would probably result in an
aolserver clone that performed quite poorly, but had good
integration support for java. It would be a huge task that wouldn't
be worth the effort. Removing the memory leaks from tclblend
would be a much easier task by comparison, and the use of
ns_javablend would allow you to easily integrate java and tcl
within aolserver.
I have nothing to say on the technical merits, of course, but I
do worry about the future for AOLServer. I worry about the fact that
90% of the developer resources for AOLServer come from AOL,
whose needs are often orthogonal to those of the OpenACS
community.
I have floated these concerns privately to a few community
members, and the answer I always get back is, "AOLServer is
still better than pretty much anything else out there. Apache
might be good someday, but it's not there yet." Etc. I understand
this argument, because it's an awful lot like the one I gave my
friends about why I was sticking with my Mac circa 1996.
I wouldn't be surprised if Apache and/or others (possibly
including some Java-based competitors) close the gap on
AOLServer on issues of performance and stability in the next
year (or two). Once they have done so, their rather large
developer base can start focusing more energy on features -- on
making web servers do new and interesting things. When that
happens, will the AOL bother to implement these new features?
Will they need them? If not, will the OpenACS community be able
to keep up? Is there any point in trying?
My suspicion is that, just as Apple found itself with nothing
competitive about its OS other than it's (user) interface,
AOLServer will someday (sooner than you think) find itself in a
position where the only thing competitive about it is the (tcl/db)
interface. If and when that day comes, we will have to do what
Apple did, i.e., rip out the guts and start over, keeping mainly a lot
of the interface stuff plus as much backward compatibility as they
could swing without killing their effort to move forward. (Please,
no flames about the OS X UI.)
If I'm right about this, then the community will probably have
to make the things you like about AOLServer portable to another
web server sooner or later. Inevitably, the first implementation
will be a huge resource drain and have poor performance
relative to the real thing. This is all the more reason to start now.
To see if I'm right, the question you have to ask is simple: Is
AOLServer evolving as quickly as the competition? If it is, then no
worries. If it's not, then no matter how far ahead it may be now, it
is inevitable that it will be surpassed, at which point the
huge investment of resources in OpenACS will start to lose
value. How fast the competition catches up --and how fast it
speeds ahead once it has caught up-- is simply a function
of how far behind the competition is and how quickly they are
catching up. Once you figure that out, then figure out how long it
would take to polish an implementation of the API on another
platform (mod_AOLServer, AOLServer J2EE,
AOLServer.NET...whatever). If the latter timeline gets a lot longer
than the former one, then the community has some serious
problems.
All this is to say that perhaps Dan's specific objections to
Jacl --as accurate as I'm sure they are-- may not be as
compelling if one takes a longer view.
