David writes:

> Nice literary reference!  Please refresh my fading memory on the origin.

The poem is the "Rubaiyat of Omar Khayyam", as translated by Edward
FitzGerald (which is by far the most commonly quoted and lyrical of all of
the translations of the Rubaiyat). The stanza of interest (which you'll see I
liberally paraphrased) is:

     The Moving Finger writes; and, having writ,
     Moves on: nor all thy Piety nor Wit
     Shall lure it back to cancel half a Line,
     Nor all thy Tears wash out a Word of it.

The Rubaiyat is a collection of a 150 Persian quatrains, written by Khayyam
(1048-1122). His full name was Hakim Abolfath Omar Ebn Ibrahim Khayyam
Neishapuri. He was the most famous Persian mathematician, engineer and poet
of his age. I've always greatly enjoyed the Rubaiyat.

As to the original question, someone else wrote me privately and asked for
more actual detail concerning the nature of the data flow in print spoolers:

> Without getting into too much detail,  what does aborting the IO do to the
> print buffer?
>
>  Will the 3000 just stop adding data to the buffer, but still try to finish
>  printing what is in the buffer and then stop the spooler?

The answer to your last question is yes, although the spooler generally quits
immediately. The question of greater interest is: What happens to the data
that the spooler has already spit out, but which has not yet printed? While I
was kidding around in my previous answer, the proper answer is: It depends.

It depends on the nature of the printer, its communication protocol, the
speed of its communication channel, and who wrote the printer drivers. A
spooler file itself is a "buffer". Once the spooler file has been released to
the printer driver, the information in the spooler is transmitted as quickly
as the printer driver will accept it. Because no printer can accept
information as quickly as a printer driver can deliver it, it has to be
buffered and flow-controlled, perhaps at several points.

There's bound to be at least a small buffer in the printer that is totally
beyond the control of the HP3000 -- and that buffer may actually be very
large. There similarly may be a buffer in the I/O communications channel that
allows for the more efficient packetization of the data stream. While
theoretically all of these buffers could be cleaned out on command from a
printer driver, in practice, it's almost never done, thus the verse tends to
be literally true.

Wirt Atmar