The Feed

5/02: The machines are all connected with 10 megbit ethernet. The UPL philosophy is that once you've added a networking topology, you never get rid of it, which explains why we have thicknet laying around the room even though nothing's attached to it.

etler: I think I cut through the drop cable for the thicknet with a bolt cutter when I gave the UPL its 100 Mbps feed circa 2000. The mangled cable is probably still under the floor of the lab opposite the UPL unless some enterprising successor to myself pulled it out.

This was not an accident. (9/03)

Don't you guys have gigabit now, at least between the CSL and a key machine or two?

etler: Wow over fiber? How'd it get in? That cable run to get from the CSL to the UPL is about the second worst I ever encountered in that god foresaken building. (The worst being the cable run over to 3310 from the CSL. Hmm... and the zeroth being the run from the CNI in the basement up to the CSL so I could get my pirated Muzak feed up there, but that's a story for some other day.)

Anyway, unless a new path was found, I pity the person who had to run fiber along that path. Category 5 UTP was enough of a bitch.

nwa: Are you sure that wasn't earlier Sam? I seem to recall the lab having 100Mb by the end of my time and I graduated in '99.. ?? (right around the time we got vger?)

etler: I'm inclined to believe that as most of those years are blurred by an alcoholic haze. I started in the CSL in Spring of 1998 I think and was gone by June of 2000. It probably wouldn't have been earlier than early 1999 though as I don't think we had readily available 100BaseTX for things like the UPL until then. Previously all high speed stuff was connected via FDDI or OC3 ATM (Cisco routers in the various IDFs mostly).

100BaseTX for the rest of the building didn't really roll out until we had the Foundry routers. At that point we did straight 100BaseTX runs to routers and switches outside of the core. The UPL had a feed directly into one of the Foundries. I think the Foundry routers showed up in 1999 but again, my memory is a bit hazy.

mturner: As of my tenure (1994-1996) there were 3 feeds coming into that room:

    1. The original thicknet (10-base5) feed (from when the UPL had moved up from the basement)
    2. A thinnet (10-base2) feed that had been used for awhile
    3. A recently-installed 10-baseT feed

Never had more than one been used at a time. Since the run was a royal pain the CSL monkeys just never bothered to remove the old cable when we got upgraded.

Internally we used all three formats:

To go with our new 10-baseT feed we had just acquired a 8-port hub with a thinnet jack. At the time 10-baseT was fairly new and hubs were still hundreds of dollars - I think we acquired ours through some "don't ask don't tell" deal with some extra hardware from some grant. Not sure about the details.

Some machines we had (notably the next cubes) only had thinnet, so we fed them off the back of the hub.

Since we STILL were perpetually out of ethernet ports we also used our original ticknet - we would chain it off the end of the thicknet (using the "evil connector" which I sure hope you've saved) This gave us constant reflection issues but by constantly tweaking (and a lot of "ping -f host > /dev/audio" sessions) we could make the whole thing more-or-less functional. After awhile what started out from neccesity became a point of ghetto pride (especially for Besh and I) and so we always lept the thicknet in the network one way or another.

These days with gigabit switches available <$100 it's hard to imagine how precious hub ports were for us.

We also had a Cayman Gatorbox routing IP from our ethernet to a localtalk network - peter (lunatic) had scavenged a bunch of localtalk and a couple old macs from his sysadmin job at the uw hospital. CSL provided us with 128.105.46/24 for our localtalk network. Localtalk is only about ISDN speed but it was good enough for running telnet sessions and maybe Mosaic/Gopher.

This was around 1995-1996.

mitch: Here's a score... an ASCII network map I made July 2, 1995 showing all the port utilization. The media-convertor is what was on port 5 of the hub (the hub's builtin thinnet was running only dax - I believe that part was relatively constant)

I don't remember what machine "spock" was at this time - its before the NeXTcubes arrived. Might have been a personal machine or something. Ditto for "tribble" - that's long before the HP machine with that name came into the lab.


Map of 128.105.45:

Key:
*** - Twisted Pair
--- - Thinnet
=== - Thicknet

         ******** chimera (21)
         *
         * ****** tribble (15)
         * *
         * * **** drizzt (17)
         * * *
         * * * ** nothing (19)
         * * * *
         _______________
         | 6 4 7 8 2|***picard (11)
         |          |
(2) gw***|1 5 T/P 3|***yar (10)
    ---------------
            |      |
            |     $+--+$
            |         |
            |        dax (16)
            |
            $+------+-------+=====+=====+===+====+======+===+$
             |      |       |     |     |   |    |      |
           eta arundel           spock      gowron      duras
           (18) (20)             (12)         (13)        (14)