craig-drexel

[ Submitted to Datafile, but I don't have a record of whether it was
ever published. ]

			     Notes On the
		  1992 HP Handheld User's Conference
			  Drexel University
			   27-29 March 1992

		      by Craig A. Finseth, #745

These notes will be incomplete as I was able to attend only about one
third of the conference.  Such is life.  Editorial comments are in []s.

I arrived in town (Philadelphia) about 10.45 on Saturday, 28 March. I
proceeded to hire a car and follow Jake's clear directions to the
site. Once there, I proceeded to get thrown way off the track by the
one-way roads.  Miss your turn and you get a scenic tour of several
kilometers of riverbank roadway before you can turn around...
Eventually, I found a car park and parked, arriving at the conference
itself about 11.30.

This was in the middle of "Using Handhelds in Mathematical Education"
by Raymond LeBarbera.  [Consult the proceedings for copies of the
papers.  I have send a copy of them to the club.]  He has undertaken
to create a comprehensive program in arithmetic drill on a variety of
programmable calculators. He supports the 48SX (ROM cards), 48S
(small, downloaded modules), 95LX (ROM cards again) and a variety of
other devices whose names will not tarnish these pages.  He has a
great enthusiasm for his work.  One audience question was especially
interesting.  It asked, in essence, "isn't the market for people who
both own 48s and are in need of arithmetic drill a little small?"
Someone else pointed out that schools may already own 48s (if only!)
and hence can now use them in more than one course of study.

Lunch was held at a "food court" a block or so away.  This was one of
those places with an array of shops around a set of common tables.
The food was good.

We started the afternoon a little late.  Unlike the recent Corvallis
conference, the schedule was plenty loose and there was no real
pressure on the speakers.

Waiting for us when we got back were some mockups of the 1152,
courtesy of Dennis York from HP (my notes list Dennis as a section
head in the R&D lab).  These were design development ideas for the
48SX and consisted of a 41-ish case with a (fake) 3-line display, a
71-ish case, a clamshell (18C-ish) case, and a case that looked a lot
like a 48.  All of these form factors were considered during the 48's
development.  In addition were a mockup of a cassette drive, and the
"innards" from a 95LX and 48SX.

I have [I hope!] arranged for pictures of these mockups and will send
a set on to the club.

Fred Linton spoke on "A Serial A/B Switch With "Double-Delta"
Extension."  His talk covered using a special type of switch that
allowed you to use a terminal to set up a transfer between two
computers and then monitor its progress.

Lew Thomas then spoke on "An HP95LX to HP-IL Interface."  He spoke at
length about what he encountered in the course of developing the
product, and his overhead slides included layouts of of the circuit
board.  He hopes to sell it for about US$200, which is impressive when
you consider that the tooling alone for the injection molded case is
almost US$20,000.  (He observed that HP had used the tooling company
first and spoiled them.)  He is very interested in hearing from users
about this product in particular or any other ideas that you might
have in general.  Copies of his survey form are included in the packet
that I shipped to the club.  Please make as many copies of the survey
as needed.

John Wettroth gave an update on "World Control," in which he described
some of the applications for his analog/digital I/O interface for the
48.  His paper covers this information well.  He also covered the
steps to follow when developing an application.  Finally, he is
considering coming out with some generic, pre-packaged sensors for use
with his interface.

After the break, Brian Maguire covered his "HP48 System RPL
Development ROM Card."  This is an ambitious project, sort of a ZENROM
or HEPAX module for the 48SX.  With it, you will be able to write,
edit, and execute System RPL routines directly from the 48SX: no IBM
PC required.  [I want one!]  His talk was a basic introduction to what
System RPL is about [see his paper and the proceedings supplement].
In his experience, System RPL programs run from 3 to 10 times faster
than user RPL and are smaller.  He estimates that this card will be
available in about two months, most of which is required to improve
the manual.

"Using UTIL Forth to Explore the HP95LX" was presented by Chris Smith.
This product is what you would expect.  As a language, Forth seems to
be well-matched to the 95: terse, low memory requirement, etc.  Of
course, you have to like Forth (:-).  Both Forth itself and programs
that you write with it are system-manager compliant.

We then adjourned to dinner, winding up at an Italian restaurant. 
People were split up among several tables.

The last item on the day's agenda was a free-form session led by
Dennis York from HP.  The nominal topic was "How Should the HP95 Be
Improved?"  However, discussion included the 48 and other devices.  I
will try to capture the main points that were brought up: don't expect
a coherent narrative.

How many of us had a 48S or 48SX?  All had the SX.

What would we like to see fixed on the 48?  The keyboard layout,
increase screen size (or at least pixels), speed up the equation
writer, faster, more memory.

Apparently, the 48's toolmaker dictated that the shift keys had to be
located around the outside edge of the keyboard.  [I assume that it
has to do with where there is enough room for the extra traces.]

Someone asked about the rationale for why things worked like they do
on the 48 (e.g., the quadratic solver is baroque when compared to the
TI or Casio versions that prompt the user).  The responses echoed both
the recent USENET News discussion (the 48 is mathematically correct
and the user may have incorrect expectations) and the 48 design
approach (the 48 provides the raw tools for you to package as you see
fit, see Wickes' 48 Insights books).

How many of us use a PC-type computer with our 48s? Most [and
therefore the serial port was a good idea].

Someone asked why there couldn't be more card slots in the 48SX.
[Other than case tooling (:-).] The answer was that the address space
was full.  Of the 512 KBytes (1 MNybble) of address space, 256 KBytes
is used for the ROM and each slot has 128 KBytes of address space
dedicated to it.  (This is also why >128 KByte cards must be
bank-switched.)  Thus, the address space is full already.

But wait!  What about the 32 KBytes of built-in RAM?  It lives on top
of the last 32 KBytes of ROM.  When code is executed from here, the 48
must first map the ROM in for reading (writes go to the RAM, of
course), copy the required code to RAM, un-map the ROM, execute the
code, and finally delete the code from RAM.  All of this entails a
20-30% performance hit overall.  But ROM was full already...

If you pull a 48 RAM card and hold it in your hand, it has about a 2
week lifetime before the battery dies.  If you put the same card into
the storage case (right away!), the lifetime increases to about a
year.  This is because the storage case incorporates a conductive
strip which keeps the leads grounded.

Most 48 users develop their code on the 48 itself (i.e., don't use
the PDL).  These are the same users who use a PC for backups, etc.

There was an involved question about a bug in the 95's MEMO
application.  The fix for this bug, of course, is to install Freyja on
top of the MEMO application. [I had to work in a plug for the
system-manager compliant version somewhere, didn't I?]

There have been no code changes in the 95 since August, even for the 1
MByte version.  [Unfortunate.]  There also is _not_ a new version of
the Hopper chip. [Doubly unfortunate, as I was looking forward to the
"hidden cursor" bug being fixed (you can't see the cursor when it is
displayed over an inverse-video space).]

Supposedly, no version F ROM will ever be released for the 48.   This
is due to the "leak" via the release of the System RPL tools last
August.  (After each version is released, the internal tools all have
their version numbers incremented to prevent confusing any code that
they affect with released code.  Someone forgot to fix this before
releasing the tools.)  There will also never be a version G ROM. 
This is because they messed things up while tinkering.  There _may_
(as in, "not impossible") be a version H ROM.

About 20% of all 95s are sold to technical / scientific users.  This
is a much larger percentage than people at HP expected.  [An article
that I read somewhere else claims that 400,000 95s were sold as of March
1992.]

He asked for changes that we wanted in the 95:

- larger display

- IN/X (a small Unix variant)

- Coherent (another operating system)

- 48 emulator

- observation that technical users use the 95 for its text and
database capabilities [observation: I have heard (and believe) that
most technical users spend about 80% of their time doing word
processing, etc., and less than 20% of their time doing technical
analyses]

Should they take something out of the 95?

- 95 without Lotus for engineers (this point was disputed: many
people like having Lotus)

- offer a symbolic math package instead of Lotus

Audience question: what percent of the (list) price is due to Lotus?
About US$100.  [My experience confirms this, I independantly estimated
the royalty to Lotus for 1-2-3 to be US$20-$25.]

- What about an RPL ROM?  It turns out that RPL doesn't run well on
Intel architectures.  It really likes the Saturn.  It could _maybe_
work on, say, a 68000-architecture machine.

- What about a 2nd CPU?  No, as there is no board space available, and
the support chips (Hopper) are build around the Intel interface
specifications. [observation: you _don't_ want to mix Intel and
Motorola chips as you will _never_ get the signalling to work
properly. Trust me on this: I tried.]

- What about selling Lotus on a ROM card? Market research shows that
the closest (#2) package sells only 1/3 the copies of Lotus.  Thus,
bundling Lotus is not an issue.

- How about a better word processor?  There are a number of word
processors being developed as ROM cards.

- HP view the current 95 as the -35A of these machines. They know that
all of the blue key applications need to be improved.  They are
looking for more radical ways to improve the product.

- There were requests for Mathematica and Quicken (I have never heard
of this one).

- How close is a 386SL version?  A long, long ways of because of power
requirements (3 years).

- Higher resolution display? No comment (Eric Vogel-style)

- What about the rest of the DOS commands?  At first Dennis simply
said that this was a great story that couldn't be told.  Eventually,
the following emerged:

	- HP looked at how much ROM space was available and made a
	list of what they wanted to include.
	- They took this list to Microsoft and licensed the code.
	- Now, when users say "what about the rest of DOS?" HP found
	that the rest wasn't included in the license.
	- So HP's and Microsoft's solicitors must get together and
	talk...

- Requests for a project scheduler and an outliner.

- People mentioned that they could do without the graphs (so that the
ROM space could be used for something else).  Dennis reported that one
of the major complaints about the machine is that it is missing the
Print Graph command.

- Someone mentioned that a bigger case would be ok, maybe 50% larger.
There were many objections to this in the audience.  Dennis mentioned
that the #1 customer like is its small size.  Also, the #1 customer
dislike is the small size.

- Request for Maple.

- Dennis offered the following suggestion: assume that ROM space is
free.  What costs are the effort, time, and royalties required for a
product.

- Someone suggested that Mathematica is not interactive enough for
the 95.

- Requests for Eureka (1 person) and T!K Solver (3 users).

- People want to carry one machine, not two.  Its size is important. 
It should be customizable.

- Don't let Lotus do your code.  Dennis responded that we should see
more HP content over time.  HP view its long term edge to be in its
software.

- HP have redone the 48 manual due to customer comments.

- On the 95, someone asked that the SERCTL and similar commands return
the current status (presumably for use in .BAT files).

- According to Dennis, the standard cell UART lost some of its
standard parts along the way to the Hopper chip.

- What about Version 2 PCMCIA?  HP are committed to the standard.
However, the standard wasn't ready for the 95 (the version 2 standard
was a year too late).

- The 1 MByte upgrade doesn't require a circuit board change.
Originally, there were 2, 512 KByte ROM chips and a 512 KByte RAM
chip.  Now that 1 MByte ROM chips are available, there is room for 2,
512 KByte RAM chips.

- Product manuals.  For example, the Casio manual (as for those of
many Japanese products) have a strong table of contents but no index.
Richard Nelson was insistent that a good index is a must.

- People preferred page numbering to section numbering.

- The biggest thing (=complaint) that Richard Nelson hears about
relates to the manual.

- On-line manuals are a possibility.

- The correct size for a manual is a subject of debate within HP
(should the manual be the same size as the product?)  The customers
feel that the manuals are too small.

- How about releasing manuals on line?  (Done for the 48 in the PDL.)

- Request for Magellan and other Lotus add-ons.

- Request for database software.

- What does "640K" mean in the 1 MByte version?  Compared to other
MS/DOS machines, there is more room left in the 95.

- Richard Nelson: when doing I/O on any HP high-end product and the
and battery dies, they don't die gracefully.

- Reversed batteries are a common problem.

- Some people think that you can reset a 48 by reversing the
batteries:

			    >>> DON'T <<<

you will tend to destroy your machine.

- In the 95's hinge, there is a clutch on the right and the left
hinge is just loose.  This is good for cradle users, as you tend to
push on the right hinge to free the 95 from the cradle.

- There is no difference in construction on 1 MByte units.  The only
differences are the two chips and the bezel printing.

- While getting production going on the cradles, a production engineer
thought that the cradle was too narrow.  It turns out that new 95s
were 3/10000" (or maybe 3/1000") bigger that spec.  Cradle was fixed.

- Cradle reduces the I/R range to 3-4 inches if using one cradle, 1
inch if using two cradles.

- The current 95 ROM version is 1.03A (Oct).  The first version was
1.01A.

- Jake Schwartz: likes the 48 because you can hold it in one hand and
operate it with the other.

- At last August's conference, Mark Smith was on the panel and
guaranteed that the hinges would work.  People are noticing
sagging and loose hinges.  This may be due to tool wear.  In any
event, there have been changes in the clutch design.

- Why doesn't the 95 open flat?  Don't know (cable flexing?
aesthetics because a joint would become visible?)

- 48 cases were being warped.  This was due to a capacitor physically
interfering with the case during assembly.  The line was shut down for
a week to find this problem.

- Paul Hubbard: It would be nice if the 95 had finger dents, as it
tends to slip out of your hand when opening it.

- The 95 is released in 11 languages.

- After marketing complained to R&D about making it easy to
internationalize the products, the 48 was designed to be fully
customizable.  Only one version (English) was released.  Such is
marketing.

Things were breaking up by 22.30, at which time I had to leave so that
I could get my hired car returned by 23.00.  Then, home on the 7.00
plane the next morning.

All in all, an excellent conference and many kudos to Jake Schwartz
and company for putting it on.

The only vendor present was the company that markets UTIL.

The notes for this article were taken using Freyja on an HP95LX (of
course!).  They amounted to several kilobytes.  Who says that you
can't type on a 95's keyboard?