[HECnet] NT 4 on AlphaServer es40

[Johnny Billquist]

On 2013-02-12 17:18, lee.gleason at comcast.net wrote:


>Speaking of IAS, it really looks cool when reading specs, but I've never
>touched it, and another aspect of those specs is that it looks like it
>would be rather slow...


    I managed an IAS system on an 11/70 back in the day. I don't have any
formal benchmarks, but it didn't "feel" any slower than equivalent
RSX11M and M+ systems I have been on.

Cool. The reason I mentioned it because of just such things as you mention below - device drivers being full blown tasks do imply quite a lot more overhead.

    IAS was a real treat. If you wanted them, you could SYSGEN in real
timesharing features for supporting users who didn't require realtime
response to their needs. It included lots of concepts   that later made
their way into VMS, (although, considering this crowd, I don't know if
that will be considered a plus here...).

Well, I for one don't mind VMS, even if I think RSX is way cooler most of the time.

    The most interesting feature for me was the way device drivers
worked. Called "handlers", they were complete tasks, instead of the APR
and a couple registers' worth of context provided by RSX11M/M+ drivers.
You could do a lot more work in them, a lot easier than on the other RSX
variants. The down side is that any driver action involved scheduling a
task rather than the lightweight context switch required by a driver.
But, having said that, the system I managed supported lots and lots of
terminals at 9600 baud, and wasn't bogged down by servicing interrupts,
so scheduling a task to do IO didn't turn out as bad as you'd think it
would.

Yeah. After having done a lot of device drivers and stuff in -11M+, I sometimes long for the freedom of a task context. That's when you go diving into ACPs, but it would be nicer if the driver was a task in itself.

    I managed RSTS sytstems as well and I vastly preferred the richer
environment provide by IAS. I recall being at the DECUS Symposium where
the future of IAS was announced (that is, no future...). There was much
lamentation, gnashing of teeth and rending of garments over that.
Support did actually continue for quite a few years after that though -
turns out that the US Air Force was a   big IAS user, and DEC didn't want
to upset the government.

    BTW, I'm always looking for IAS related "stuff" - copies of the
DeVIAS newsletter, IAS software (espcially DECnet-IAS) and the like.

I believe IAS never went to Mentec, but actually stayed with DEC, for the very same reason (although I also heard that supposedly it was IRS, but urban legends are probably abundant).

>Never seen TRAX in real life either, btw. What was so good about it?

    Me neither, though I did use some of the VT61 and VT62 terminals that
were developed for it - like VT-52s, but with IBM style block mode. Not
a lot of software outside of TRAX used their block mode features, so
they were pretty rare.

The VT61... The VT62 was a different beast, and I had one for many years... It's basically a VT52 with inverse video attribute.

	Johnny

-- 
Johnny Billquist                                   || "I'm on a bus
                                                                  ||   on a psychedelic trip
email: bqt at softjar.se                         ||   Reading murder books
pdp is alive!                                         ||   tryin' to stay hip" - B. Idol