[Prev][Next][Index][Thread]
Experimental TCP Interface replaces MacTCP Mac CL-HTTP 58.12
I have placed an experimental version of a general overhaul of the MacTCP file
on the FTP site in
ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/devo/xmactcp-1-0.sea.hqx
This revision aims to fix all lossage related to CL-HTTP operation on the
Mac under MCL 3.0.
It has been tested under MacOS 7.5.3 running Open Transport 1.1 on a 9500.
If you don't
already have 7.5.3 and OT 1.1, you can get if from the following url
without waiting forever.
ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/macos-7-5-2.sea.bin
Causal observation suggests that CL-HTTP on the 9500 is substantially
faster with this configuration.
Some work on MCL and the TCP interface may be needed to reduce latency further.
Interested parties are encouraged to test this out and report any crashes
or other shortcomings
to bug-mac-cl-http@wilson and bug-mcl@digitool.com. Better yet, send back
the fixes since
nobody is paid to do this.
Once we are convinced that this really works across the board and it is a
better bet than the 58.12
status quo, we'll incorporate this as the default. But, your feedback will
be needed to ensure we have
the bugs out. Drop a note to bug-mac-cl-http describing all successes and
be sure to include the
configuration that you are running.
I wrote most of this on Saturday for fun because I was sick of the MCL
losing and making CL-HTTP look bad.
I read the MacTCP manual, looked at RJH's hack, and got some important
insight and help from Kalman Reti.
It turns out that RJH's code misses the mark in a number of ways, even if
avoiding parameter block collisions
and resourcing connections are useful concepts. Nevertheless, it did
provide the proximate call to action.