stacked disciplines for i/o devices

[trb at masscomp.UUCP]

The idiot in me insists that I continue to feed the flames:
How can I resist responding to a personal affront?

William LeFebvre at Rice says:

	Given a typical system, there are a wider variety of terminals
	than disk drives.  Mainly because there are about 20 times more
	terminals on the system than disk drives.  Also, given a
	typical system, terminal types tend to change more often than
	disk drive types.  Mainly because there is an order of
	magnitude complexity difference between changing a terminal and
	changing a disk drive.  Not to mention the fact that people who
	dial up a computer use whatever terminal they have at home.
	When was the last time you changed the type of drive you are
	using?  When was the last time you added or swapped out a
	terminal?  When was the last time someone connected a disk
	drive to your system over a phone?  Your argument is fairly
	vacuous.  It's quite true that the disk drive characteristics
	are hard coded in the kernel.  But they don't change very
	often.  You can afford to hard code disk drive
	characteristics.  I'm not saying it's a good idea, but you can
	get away with it.  You CAN'T afford to do that with terminals!
	And anyone that does so is painting himself into a corner.

William,
In your lead statement, you make a false assumtion about what a
typical system is, and you also make a misleading statement about
comparative numbers of terminals and disks.  You say that there are
20x more ttys than disks.  I don't know what your idea of a typical
system is, but the typical system I have sitting next to me has one
terminal and two types of disk on it.  Your typical system might have
twenty times as many terminals as disks, but I'd be surprised if it
had twenty times as many different kinds of terminals as kinds of
disks.  Yes, I know that termcap supports a jillion kinds of
terminals, but I dont know of a single installation that uses more than
oh, twenty different kinds.  Most use about five different kinds.

Did you ever wonder why people use many kinds of terminals but not many
kinds of disks?  Perhaps it's because the software support is there (!).
In the Bell System (where I had knowledge of the configuration of
hundreds of systems), UNIX users have been stuck with certain hardware
vendors even though the price and performance were certainly not
competitive with current technology, all because of questionable
software and hardware support worries.

Judging by your answer, I can only assume that you haven't hacked up
too many disk drivers to run with difference sizes or partitions.  I
haven't hacked up very many, I've hacked up too many.  On my typical
system, when I change my terminal, I don't really notice it because I
just diddle my terminal name and termcap takes care of the rest.  When
I have to bring up a new flavor of disk, I have to noodle around in the
damned driver.  Right now, here at Masscomp, we have everything from
floppies to 5.25" winnies to 8", 10", and 14" drives in various heights
and girths.  It's our business to stay abreast of current storage
technologies, and we see a lot of different configurations fly by.  Or,
to contradict the way you put it, they change very often.  We have
many more kinds of disk drives around here than we have kinds of
terminals, because disk technology is changing faster than terminal
technolgy.

You talk about what you can "afford to do" and what's a "good idea."
You also called my argument of having easily configurable stacked disk
i/o disciplines "fairly vacuous."  My argument was either valid or
not.  I am either vacuous or not.  Realize that "vacuous" is just a
snot-nosed, recherche, euphemism for "stupid." After you called my idea
vacuous, you presented the hard coding alternative this way: "I'm not
saying it's a good idea, but you can get away with it."  I'm not
convinced.  In this case, I'd say that my argument is valid, in MANY
instances.  There are lots of people out here who need to be able to
integrate new hardware technology into their systems.  Just because you
don't, isn't sufficient reason to call us vacuous.

Remember, kids, i/o devices all perform pretty much the same function.
Classes of hardware may be grouped such that the hardware dependent
parts are isolated, and the hardware independent parts are serviced by
common software.  This goes for disks and tapes, as well as we have
learned that it goes for TTYs and CPUs.

	That's one to grow on.
		- Mr. T

	I pity the fool who calls me vacuous.
		- also Mr. T

	Andy Tannenbaum   Masscomp Inc  Westford MA   (617) 692-6200 x274