Loadstar 246/246.d81/t.qlink aol

Text File | 2004-01-01 | 5.8 KB | 207 lines

u
        The QLINK/AOL "Sogga"
           by Randell Jesup


This is a reply to an ancient message
found via DejaNews.

Brian Heyboer writes:

>Brian Williams wrote:

>> I've been reading about this for
>> some time, and.. can anyone tell me
>> WHY aMERICA oNLINE won't sell the
>> software or the rights to run
>> Qlink??  What the hell are they
>> scared of?!

>
> I can't tell you for sure, but
> I suspect they are afraid it will
> give away some of their security
> systems that are also used in the
> AOL software.  Remember, there was a
> lot more on the Q-Link end than just
> the interface for the users.  There
> was also their entire billing and
> password security system. There was
> also a "back door" of sorts where
> Q-Link menus and what-not could be
> updated via AOL.
>

    AOL is in fact largely based on
rewritten QLink (nee PlayNet) code.
Many of the algorithms are unchanged.

>
> Another possibility is that they
> cannot rather than will not.  Q-Link
> licensed the software from Playnet
> and acquired the rights to it only
> after winning a lawsuit against the
> receiver of the bankrupt Playnet.
> They never did get all the source
> code and documentation the lawsuit
> gave them the rights to.  So, they
> may not be able to either because
> the terms of the judgement don't
> allow it or they simply don't have
> it all.

    In fact, they may not have the
right to resell the technology; it
depends on what rights they got. (I
suspect they eventually got all the
rights, though.)

    They did, however, have all the
source code and documentation for the
PlayNet system, at least as it was
when they licensed it (we made a
number of mods later to PlayNet, some
of which were activated and some never
were). I spent a number of days down
there training various programmers
there on the design.

    One thing added after QLink (now
AOL) licensed PlayNet was a quite
complete auditorium/ panel/ etc setup
with queuing, moderators, etc, run
entirely via online messages (no
client software change required). This
was complete and tested and finished
the week before PlayNet declared
bankruptcy, so no one ever actually
used it. There were other things too,
but I remember that because I was
working on it as PlayNet went under.
Of course, they made their own mods
(initially mostly cosmetic, but they
added lots of stuff later).

    As must be obvious, I was one of
the main (and last) programmers at
PlayNet. It's [really] amusing to look
at AOL today and say, "I know why
users are limited to 10-character
names.", and see many other elements
of the original PlayNet design
unchanged (even though the reason for
them is LONG gone).

    For example, the 10-character name
limit was largely based on how many
screen names we could display in the
room header in chat within 4(?)
40-character lines on a C64 screen.
Ditto the screen-name defaults (I
remember us sitting around BS'ing
about how we'd handle that, and
conflicts -- so now you have
JoeS12345.) Online messages and how
they popped up were another Playnet
idea (remember, the next-most-
sophisticated system at the time was
Compuserve's ASCII "CB". Much has
changed in AOL, of course, but it's
kind-of heartening to see just how
well a design from 1984-85 for 64K
6502-based machines has held up over
the years, at least in the broad
strokes.

    The system (PlayNet and QLink) was
actually quite sophisticated. It was
run by programs written in a
multi-tasking state-machine language.
(Yes, your C64 was multi-tasking when
doing this - N state-machine tasks
plus the "main" (basic/etc) task,
which ran the game or whatever if
needed. Things like Online messages
caused a new task to be started.)

    The communications protocol was
designed (by me) to error-correct the
X.25 pad<->modem link, obey a limit on
packet size (128?), and minimize the
number of packets (since we were
charged both by the hour and the
packet back then). It used CRC
error-checking (yes, in a C64),
asymmetric sliding-windows,
piggybacked-acks, selective
retransmit, etc.

    It may be that this protocol
continued (continues?) to be used in
modified form in AOL, from what an AOL
engineer told me shortly after
QuantumLink launched AOL. I also wrote
the fast-loader (which hid itself
under the screen ram when not in use,
and used huffman compression of all
files to help speed loads) and other
bits like the server side of the BBS
(news-sort-of) section (my original
rejected design was much more like
News; amusing since I hadn't seen News
at that point).

    The server side ran (and runs)
under Stratus VOS on Stratus fault-
tolerant hardware (originally at
PlayNet a Stratus 200, with 8 12Mhz
68010's). The server-side software
was (and may still be) written in PL/1
subset G, as was most of Stratus
software of the era, including the OS
(C was introduced around the time
PlayNet went under).

    The design of the Playnet server
software was specifically set up to
make maximal use of multiple servers
connected by medium-speed links. For
example, each chat room was a separate
process (if I remember correctly, or
maybe each process handled N rooms),
and all the IPC was set up to use
Virtual Circuits (sort-of equivalent
to sockets), so they didn't care
whether the other end was on the same
system or not.

    AOL still runs on Stratus
hardware, and pushes it (and VC's etc)
to the limit I'm told. For a while
they kept an old Stratus in the corner
of the machine room for C64 owners, as
I'm sure you know.

    It's also amusing to now see the
explosion of networked games; many
ideas just now reaching the public are
very similar to the stuff we built
prototypes of or sat around discussing
back then.

    If people really want, I suppose I
could write up my view on the soap
opera of how PlayNet was, what
happened, how AOL nee QLink nee CVC
got the software, etc.

--
Randell Jesup, Scala US R&D,
Ex-Commodore-Amiga Engineer class of
'94

[e-mail]