Missing Source and Transcript problem

Two problems in one here!

Evaluating ..

Class allClasses asSortedCollection do: [:eC |
    eC selectors do: [:eS |
        (eC compiledMethodAt: eS) getSource = '"Source unavailable for this
method"'
            ifTrue: [Transcript print: eC; tab; print: eS; cr]]]

.. shows a lot of methods with unavailable source. I don't think it's
intentional is it?

Secondly, the Transcript doesn't display all the Class/Selector pairs that
fail the above test as it seems to fill up after a certain number.
Similarly, evaluating

1 to: 10000 do: [:i | Transcript print: i; cr]

stops displaying after 5183.

Using XP home if that makes a difference?

Regards
    Ian

Ian

You wrote in message news:[hidden email]...
> Two problems in one here!
>
> Evaluating ..
>
> Class allClasses asSortedCollection do: [:eC |
>     eC selectors do: [:eS |
>         (eC compiledMethodAt: eS) getSource = '"Source unavailable for
this
> method"'
>             ifTrue: [Transcript print: eC; tab; print: eS; cr]]]
>
> .. shows a lot of methods with unavailable source. I don't think it's
> intentional is it?

Hmmm, not really. A limitation of the unlock mechanism means that it is
treating the beta distrib. as an evaluation copy and removing a lot of
source. Obviously this limits the usefulness of the beta somewhat (!). Andy
and I will discuss and see what we can do.

> Secondly, the Transcript doesn't display all the Class/Selector pairs that
> fail the above test as it seems to fill up after a certain number.
> Similarly, evaluating
>
> 1 to: 10000 do: [:i | Transcript print: i; cr]
>
> stops displaying after 5183.
>
> Using XP home if that makes a difference?

Actually I noticed something similar to this yesterday while testing my
DolphinWiki restore class (also !). The Transcript is now using a
MultilineTextEdit because this avoids the some of the bugs that exist in the
RichTextEdit (e.g. output to a minimized Transcript in 4.01 cannot be seen
until one resizes the view). However it also seems to limit the amount of
text displayed to 32Kb by default on XP. This same issue has always existed
on the RTE (see for example RichTextEdit>>setMaxTextLimit and its senders).
I'm not sure at the moment whether this has always been the case, or is just
an XP thing. Anyway it is logged as defect # 456, thanks.

I suppose, on reflection, that we can't use a MultilineTextEdit because that
has limitations on Windows 98/ME anyway. Perhaps this is a good point to
test the water by mentioning that it is our intention that D5 will be the
last version of Dolphin that we will support on Windows 9X. This will
release us from the multitude of incompatibilities between the 9X and NT
lines, and allow us to start making use of some of the more interesting new
Win32 APIs. Also we will not officially be supporting D5 on Windows 95 - it
may work but we will not be doing any testing on that platform. The solution
for Windows 95 will remain D4.

Regards

Blair

Hi Blair,

I'm just getting ready to install, but, might as well comment on the Win9x
issues:

> I suppose, on reflection, that we can't use a MultilineTextEdit because
that
> has limitations on Windows 98/ME anyway. Perhaps this is a good point to
> test the water by mentioning that it is our intention that D5 will be the
> last version of Dolphin that we will support on Windows 9X.

Just guessing, D6 should appear about a year from now???  That's still
within reasonable service life of $10,000+ worth of machines that we bought
with Win98.  Upgrading them might not be an option.  I've just finished (I
hope!!) configuring my Win2k based server and should have the first Win2k
pen tablet on an evaluation basis fairly soon.  If 2k helps with
reliability, I'll argue to get rid of (or at least upgrade if practical) the
9x boxes ASAP.  Failing that, it could be a problem.  Also, ten grand is
_nothing_ compared to the hospital's installed base; fortunately a lot of
that is NT4.  Still, 9x won't go away as easily as one might like.


> Also we will not officially be supporting D5 on Windows 95 - it
> may work but we will not be doing any testing on that platform. The
solution
> for Windows 95 will remain D4.

I'll give you that one :)  However, we're about to find out whether or not
it works =:0

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Bill

You wrote in message news:[hidden email]...
> Hi Blair,
>
> I'm just getting ready to install, but, might as well comment on the Win9x
> issues:
>
> > I suppose, on reflection, that we can't use a MultilineTextEdit because
> that
> > has limitations on Windows 98/ME anyway. Perhaps this is a good point to
> > test the water by mentioning that it is our intention that D5 will be
the
> > last version of Dolphin that we will support on Windows 9X.
>
> Just guessing, D6 should appear about a year from now???  That's still
> within reasonable service life of $10,000+ worth of machines that we
bought
> with Win98.  Upgrading them might not be an option.

Upgrading them may not be an option, but as I said, Dolphin 5 (and indeed 4)
will still be around to support Windows 9X.

Since you bring up the subject of finance (how many machines does $10,000
actually represent?), the main reason to consider dropping support for 9X
(apart from being free to use cool new stuff) is the question of resources.
As a small organisation we simply don't have the time to implement and test
solutions for an increasing range of OSs. The NT line (and in particular 2K
and XP) exhibit a high degree of compatibility, and by concentrating on that
we can spend much less time on niggling compatibility issues (and working
around bugs and limitations in the 9X OSs), and much more time on moving the
product forward, improving the documentation, etc.

In all probability app's deployed from D6 will be runnable on 98 (though it
is really too early to say), but we won't be paying any attention to making
the IDE run on that platform, especially as we believe it is unsuitable for
"development" anyway.

>...  I've just finished (I
> hope!!) configuring my Win2k based server and should have the first Win2k
> pen tablet on an evaluation basis fairly soon.  If 2k helps with
> reliability, I'll argue to get rid of (or at least upgrade if practical)
the
> 9x boxes ASAP.  Failing that, it could be a problem.  Also, ten grand is
> _nothing_ compared to the hospital's installed base; fortunately a lot of
> that is NT4.  Still, 9x won't go away as easily as one might like.
>

True. Unfortunately Microsoft lengthened its life with a change of tack for
Win2k (which was originally supposed, like XP, to supercede both NT and 9X),
and then ME. Not a lot of people may be using ME, but the fact that it
appeared adds a year or two to the life of the 9X product line.

>
> > Also we will not officially be supporting D5 on Windows 95 - it
> > may work but we will not be doing any testing on that platform. The
> solution
> > for Windows 95 will remain D4.
>
> I'll give you that one :)  However, we're about to find out whether or not
> it works =:0

Good luck

Regards

Blair

Hi Blair,

>
> In all probability app's deployed from D6 will be runnable on 98 (though
it
> is really too early to say), but we won't be paying any attention to
making
> the IDE run on that platform, especially as we believe it is unsuitable
for
> "development" anyway.

I agree for professional development, but to be honest, I see Dolphin as a
serious competitor of the VB market as well, which is probably the most used
'home development' environment. People that want to try Smalltalk typically
'play' with it at home for a while before trying to convince their
management that they've found an environment with serious productivity gain
(right Bill?). And of course they pick up Dolphin as that is the easiest to
use Smalltalk version around.

>
> >...  I've just finished (I
> > hope!!) configuring my Win2k based server and should have the first
Win2k
> > pen tablet on an evaluation basis fairly soon.  If 2k helps with
> > reliability, I'll argue to get rid of (or at least upgrade if practical)
> the
> > 9x boxes ASAP.  Failing that, it could be a problem.  Also, ten grand is
> > _nothing_ compared to the hospital's installed base; fortunately a lot
of
> > that is NT4.  Still, 9x won't go away as easily as one might like.
> >
>
> True. Unfortunately Microsoft lengthened its life with a change of tack
for
> Win2k (which was originally supposed, like XP, to supercede both NT and
9X),
> and then ME. Not a lot of people may be using ME, but the fact that it
> appeared adds a year or two to the life of the 9X product line.

ME falls in the 9x line, doesn't it? So basically all people that bought a
pc for home use up to autumn 2001. People typically replace their pc after
5? years, so when D6 comes out in 2003, then 3/5 of the home users will have
a 9x machine.... (they can of course just buy my book and use the free D4VE
that comes with it ;-))

Ted

Hi Blair,

> Upgrading them may not be an option, but as I said, Dolphin 5 (and indeed
4)
> will still be around to support Windows 9X.

I'm very interested in seeing whether D5 fixes a problem I've been having
that _could_ be related to the problem of D4 (and 3???) yielding too easily
when running against a CPU hog.  The point being that image problems I can
fix; VM problems require a patch from you, and that gets harder to obtain as
time goes on and you're supporting say D6 and working on D7.


> Since you bring up the subject of finance (how many machines does $10,000
> actually represent?),

Only 3, which on one leve increases the insult of trashing them.  If there's
a "2k is demonstrably more stable" argument, then I'll happily make that.
Still, it's ten grand worth of hardware bought for a pilot project that's
growing.  But keep in mind the (guessing) >$100k worth of affected machines
owned by my department and our hospital.  Being unable to target them could
be a big problem some day.


> the main reason to consider dropping support for 9X
> (apart from being free to use cool new stuff)

In theory at least, one can take advantage of things when available w/o
requiring them.


>  is the question of resources.
> As a small organisation we simply don't have the time to implement and
test
> solutions for an increasing range of OSs. The NT line (and in particular
2K
> and XP) exhibit a high degree of compatibility, and by concentrating on
that
> we can spend much less time on niggling compatibility issues (and working
> around bugs and limitations in the 9X OSs), and much more time on moving
the
> product forward, improving the documentation, etc.

Understood.  But there are a LOT of 9x machines in service, and in part
thanks to tax laws, they'll be running for quite a while.  Look how long it
took Delta Airlines to replace their terminals; they waited because it was
_really_ expensive to replace that many machines.


> In all probability app's deployed from D6 will be runnable on 98 (though
it
> is really too early to say), but we won't be paying any attention to
making
> the IDE run on that platform, especially as we believe it is unsuitable
for
> "development" anyway.

My problem with developing on a more stable platform than my users have is
that they find the bugs instead of me.  I often use nicer machines to do
certain kind of programming (no need to fight the box when I'm refactoring
or editing a big view resource), but, I test and do a fair amount of
development on junk that approximates what my users have to tolerate.


> True. Unfortunately Microsoft lengthened its life with a change of tack
for
> Win2k (which was originally supposed, like XP, to supercede both NT and
9X),
> and then ME. Not a lot of people may be using ME, but the fact that it
> appeared adds a year or two to the life of the 9X product line.

Which only strengthens the argument for running on 98.  If the IDE won't
run, debugging 98 specific problems would be difficult at best.


> > > Also we will not officially be supporting D5 on Windows 95 - it
> > > may work but we will not be doing any testing on that platform. The
> > solution
> > > for Windows 95 will remain D4.
> >
> > I'll give you that one :)  However, we're about to find out whether or
not
> > it works =:0
>
> Good luck

FWIW, it seems to work.

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Ted,

> I agree for professional development, but to be honest, I see Dolphin as a
> serious competitor of the VB market as well, which is probably the most
used
> 'home development' environment. People that want to try Smalltalk
typically
> 'play' with it at home for a while before trying to convince their
> management that they've found an environment with serious productivity
gain
> (right Bill?).

Yup.  It's even more important because said management will look for any
excuse to dismiss "that language" before bothering to learn anything about
it.  Premature clobbering of 9x would be a ready-made excuse.


> ME falls in the 9x line, doesn't it? So basically all people that bought a
> pc for home use up to autumn 2001. People typically replace their pc after
> 5? years,

Businesses often keep them that long because of idiotic tax laws.  I knew of
one company that had two sets of "books" for computers; the realistic value
of the machine for their planning purposes, and the value that IRS made them
claim - guess who got the mine and who got the shaft.  That organization
isn't doing as well as it was, so my hunch is they have an increasing number
of just plain old machines in use.


> so when D6 comes out in 2003, then 3/5 of the home users will have
> a 9x machine.... (they can of course just buy my book and use the free
D4VE
> that comes with it ;-))

Some of them might upgrade the OS.  I might nudge the numbers down a bit,
but, it's still a lot of machines and would-be users that would be cut off.

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

"Blair McGlashan" <[hidden email]> wrote in message
news:[hidden email]...
> has limitations on Windows 98/ME anyway. Perhaps this is a good point to
> test the water by mentioning that it is our intention that D5 will be the
> last version of Dolphin that we will support on Windows 9X. This will

ho-hum. While I understand chores of supporting 9x, and financial
constraints for doing so, I still think this would be premature. Luckily my
members are scheduled for new workstations this year, but if they were not,
I would probaly be in serious trouble. Also I think of writing some
shareware with Dolphin; cuting out 9x customers seems very risky.

But, I guess there is still some time to discuss the matter, we are only
testing D5 today :)

Davorin Rusevljan

Bill,

snip

> My problem with developing on a more stable platform than my users have is
> that they find the bugs instead of me.  I often use nicer machines to do
> certain kind of programming (no need to fight the box when I'm refactoring
> or editing a big view resource), but, I test and do a fair amount of
> development on junk that approximates what my users have to tolerate.

I may be missing something here but surely in many cases it's possible to
upgrade the machines' OS  to, say, Win XP rather than junking them. I'm
running an old DELL P350 with 128Mb RAM. It was originally a P200 with 64Mb
bought in 1997. Since then it was hardware upgraded to it's current state in
1999 and now to have Windows XP installed in 2002. The thing runs fine. In
fact I was actually surprised that it boots into XP faster than a much newer
P2-800!

Best Regards,

Andy Bower
Dolphin Support
http://www.object-arts.com
---
Are you trying too hard?
http://www.object-arts.com/Relax.htm
---

Davorin

"rush" <[hidden email]> wrote in message
news:[hidden email]...
>
> "Blair McGlashan" <[hidden email]> wrote in message
> news:[hidden email]...
> > has limitations on Windows 98/ME anyway. Perhaps this is a good point to
> > test the water by mentioning that it is our intention that D5 will be
the
> > last version of Dolphin that we will support on Windows 9X. This will
>
> ho-hum. While I understand chores of supporting 9x, and financial
> constraints for doing so, I still think this would be premature. Luckily
my
> members are scheduled for new workstations this year, but if they were
not,
> I would probaly be in serious trouble. Also I think of writing some
> shareware with Dolphin; cuting out 9x customers seems very risky.
>
> But, I guess there is still some time to discuss the matter, we are only
> testing D5 today :)

Yes, and this is the whole point. 9x will continue to be supported by D5.

At this stage our feeling is that we don't want to constrain D6 to what is
possible on 9x - as we've often said in the past, it is not our intent to
product a lowest common denominator solution, and this applies just as much
within the (currently very broad) Windows family, as it does to
cross-platform stuff. The point has come (or probably was passed a long time
ago) that supporting 9x does significantly constrain what one can do, both
in terms of development effort and features. Microsoft get around this
problem by effectively upgrading the operating system by insisting that one
install some particular version of Internet Explorer (and in fact we have
made the same stipulation ourselves), but this only works up to a point, and
thus ME was the last of the line.

Regards

Blair

Andy,

> > My problem with developing on a more stable platform than my users have
is
> > that they find the bugs instead of me.  I often use nicer machines to do
> > certain kind of programming (no need to fight the box when I'm
refactoring
> > or editing a big view resource), but, I test and do a fair amount of
> > development on junk that approximates what my users have to tolerate.
>
> I may be missing something here

The problem is that the machines are pen tablets.  There are interesting
logistic issues as well as drivers to break.  On the other extreme, a
would-be customer with hundreds of computers will probably take a dim view
of upgrading their "fleet" just to run my Dolphin-based app; I would simply
be rejected as a vendor.


> but surely in many cases it's possible to
> upgrade the machines' OS  to, say, Win XP rather than junking them. I'm
> running an old DELL P350 with 128Mb RAM. It was originally a P200 with
64Mb
> bought in 1997. Since then it was hardware upgraded to it's current state
in
> 1999 and now to have Windows XP installed in 2002. The thing runs fine. In
> fact I was actually surprised that it boots into XP faster than a much
newer
> P2-800!

Quite true.  In fact, I bought a copy of 2k just before XP became commonly
available with just that in mind.  It's not an issue to do on a a machine or
two that I use.  It probably is a big deal on the pen tablets (technical
issues, not money).  It's another thing entirely to tell a would-be customer
that they have to upgrade because my "compiler" (don't get me started<g>)
won't target their machines.

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Blair,

> > But, I guess there is still some time to discuss the matter, we are only
> > testing D5 today :)
>
> Yes, and this is the whole point. 9x will continue to be supported by D5.

Understood.  But what about back-propagating VM fixes?  Part of my concern
is that something like the yield bug fix wouldn't make it into D5 from D6
(in fairness, it didn't make it into D4, perhaps in part because nobody
asked for it??).  If D5 is the terminal 9x support and you are willing to
update the VM if necessary, then we might be ok.  What do you think?

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Ian Bartholomew <[hidden email]> wrote...

> Evaluating ..
>
> Class allClasses asSortedCollection do: [:eC |
>     eC selectors do: [:eS |
>         (eC compiledMethodAt: eS) getSource = '"Source unavailable for
this
> method"'
>             ifTrue: [Transcript print: eC; tab; print: eS; cr]]]
>
> .. shows a lot of methods with unavailable source. I don't think it's
> intentional is it?

Same here - seems to be mostly (all?) non-Value Edition features (Database
Connection, Sockets, XML...).

Other than that - looks fantastic!

John Aspinall
Solutions Software
http://www.solutionsoft.co.uk/restore

Bill

You wrote in message news:[hidden email]...
>
> > Upgrading them may not be an option, but as I said, Dolphin 5 (and
indeed
> 4)
> > will still be around to support Windows 9X.
>
> I'm very interested in seeing whether D5 fixes a problem I've been having
> that _could_ be related to the problem of D4 (and 3???) yielding too
easily
> when running against a CPU hog.  The point being that image problems I can
> fix; VM problems require a patch from you,

D5 represents the culmination (probably) of Dolphin versions on its
particular VM technology. D4 has had only minor VM issues identified and is
a very stable platform. We haven't issued a patched VM for D4 for many
months. D5s VM is deliberately very little different from D4s. There may be
problems with it but we would certainly hope to identify and correct these
before the release of D6.

As for the "too friendly to CPU hogs" problem, it was not at all clear that
this was in fact a VM issue, and that wasn't discovered until we looked into
it in earnest. As this wasn't a particularly high priority bug (if one
grades bugs from 'showstopper' through 'crash that can be worked around',
etc), we didn't actually investigate it thoroughly until fairly recently,
although we had analysed it briefly before to see if it was a
straightforward problem. If we'd known it was a VM "bug" earlier, we'd
probably have released a patch for it, since we generally treat VM issues
with a higher priority for the reason you mention. In fact it appeared to be
some change in behaviour in the image.

>...and that gets harder to obtain as
> time goes on and you're supporting say D6 and working on D7.

Free support will obviously diminish. Paid for support can always be used to
get a higher service level, or in cases where your organisations priorities
are different to OAs.

>
> > Since you bring up the subject of finance (how many machines does
$10,000
> > actually represent?),
>
> Only 3, which on one leve increases the insult of trashing them.

The point of my question is not to make an insult, but to assess the expense
and effort involved in upgrading the OS. If you are prepared to upgrade from
one version of a development system to the next (D5 to D6), then presumably
upgrading the OS is not out of the question. Any sort of software upgrade
like this will involve risk, so in fact the lowest risk option would be not
to continue to use D5 to maintain those systems. I realize there are other
considerations (like how do you then take advantage of new features in D6,
and maybe develop new systems in D6), but if one is prepared to upgrade one
piece of software, then why not another? Well of course there is the cost,
if the number of machines involved is small, then the cost is not high,
especially for recently purchased machines with adequate resources.

>...If there's
> a "2k is demonstrably more stable" argument, then I'll happily make that.

Well, there can be absolutely no doubt about that. For some time Microsoft
have marketed 9x as being for home, rather than commercial use, with NT/2k
being appropriate for the latter. For once the marketing is not total
blather.

> Still, it's ten grand worth of hardware bought for a pilot project that's
> growing.  But keep in mind the (guessing) >$100k worth of affected
machines
> owned by my department and our hospital.  Being unable to target them
could
> be a big problem some day.
>
>
> > the main reason to consider dropping support for 9X
> > (apart from being free to use cool new stuff)
>
> In theory at least, one can take advantage of things when available w/o
> requiring them.

Up to a point that is possible, but sometimes it is not workable or just too
expensive to do, particularly if it is "core". A case in point is Unicode.
Internationalisation of Dolphin would be significantly easier if we just
adopted it wholesale. However we'd then have had to emulate the majority of
the Unicode versions of the Win32 API functions, a significant undertaking.
I seem to remember reading that Microsoft have, very recently, implemented
an emulation library themselves, but it has taken them years to do so, and
it seems unlikely that they will migrate other new technologies back to
Windows 9x.

> Understood.  But there are a LOT of 9x machines in service, and in part
> thanks to tax laws, they'll be running for quite a while.  Look how long
it
> took Delta Airlines to replace their terminals; they waited because it was
> _really_ expensive to replace that many machines.
>

There are probably still a lot of Windows 95 machines in service, but
Microsoft themselves no longer support it. They will no longer support
(apart from online support) Windows 98 as of June 30, 2003. Note also that
98 enters the "Extended" support phase at the end of June this year, which
means they downgrade the level of free support they provide (I think).

Smalltalkers in general (including us) probably rely too much on the fact
that we are running in a live environment, literally treading on our own
toes sometimes, to test our stuff. We ourselves have certainly done this in
the past, and still do (though to a lesser extent). Really though it is no
substitute for proper testing. I get a much more comfortable feeling when my
SUnit tests continue to pass after a major change after a run of a few
minutes, than I do from days of bashing away with classes that might or
might not exercise that change. Yes you have to test in the production
environment, but it makes sense to develop in the most productive one. Even
just the time saved from not having to reboot 5x/day raises one's
productivity a lot :-). A specific problem on 9x is that the resource
limitations are just too tight for a development environment like Dolphin,
and one is constantly living on the edge. When I ran Windows 98 on my laptop
I used to get all sorts of unexplained crashes that I never saw on WinNT. I
eventually discovered (by running a resource meter all the time) that most
of these, and a lot of other odd behaviour, was due to running out of
resources. This was exacerbated by my development style, which involves
opening loads and loads of windows, but even so I think most users would run
into this fairly frequently. If I'd only ever run Dolphin on there, I would
have been disappointed by its stability.

In summary: We haven't made a specific decision not to support 9x in the
next version of Dolphin. We haven't necessarily identified any particular
technologies that would prevent us running on 9x in the future, but at this
stage we don't want to constrain future products purely for that reason. At
some point we will have to make a decision not to support 9x in a new
release. It will sure as hell be earlier than five years away (Microsoft
will probably drop support before then anyway, just as they no longer
support Windows 95), and could well be in the next major release. When we do
that we will continue to support the last 9x compatible release, within
reason.

Thanks for your (and everyone elses') feedback on this matter. We will
obviously take it all into consideration.

Regards

Blair

Hi Blair,

> D5 represents the culmination (probably) of Dolphin versions on its
> particular VM technology.

That opens up another can of worms, if you mean what I think you mean.  .NET
might force me to "hang back", unless you can find a way to support
resumable exceptions on the UVM, or if I can find a way to live without
them.  I'd probably have to talk you into some changes to the base system -
we can discuss that if the time comes.  As you know, I see little good
coming from turning the VM over to MS.  Hopefully either I'm wrong, or MS
will discover early on that I'm right about the revenue it won't generate
and reverse course before they do too much damage to the market.


> As for the "too friendly to CPU hogs" problem, it was not at all clear
that
> this was in fact a VM issue, and that wasn't discovered until we looked
into
> it in earnest. As this wasn't a particularly high priority bug (if one
> grades bugs from 'showstopper' through 'crash that can be worked around',
> etc), we didn't actually investigate it thoroughly until fairly recently,
> although we had analysed it briefly before to see if it was a
> straightforward problem. If we'd known it was a VM "bug" earlier, we'd
> probably have released a patch for it, since we generally treat VM issues
> with a higher priority for the reason you mention. In fact it appeared to
be
> some change in behaviour in the image.

Fair enough.  I'm looking forward to having the fix because it will at a
minimum rule out the bug as the cause of a rare but severe problem.  I'm
also suspicious that it had a role in the problems I had with the overlapped
call to copy files.


> Free support will obviously diminish. Paid for support can always be used
to
> get a higher service level, or in cases where your organisations
priorities
> are different to OAs.

That's certainly reasonable.


> > Only 3, which on one leve increases the insult of trashing them.
>
> The point of my question is not to make an insult,

Please understand how I meant 'insult' - financially, not implying any
offense taken.


> but to assess the expense
> and effort involved in upgrading the OS.

Realistically, I think it's toss and replace, but, I would be pleasantly
surprised to discover that it's otherwise.


> If you are prepared to upgrade from
> one version of a development system to the next (D5 to D6), then
presumably
> upgrading the OS is not out of the question. Any sort of software upgrade
> like this will involve risk, so in fact the lowest risk option would be
not
> to continue to use D5 to maintain those systems. I realize there are other
> considerations (like how do you then take advantage of new features in D6,
> and maybe develop new systems in D6), but if one is prepared to upgrade
one
> piece of software, then why not another?

I'm very open to that when I have control.  One of my points is that we as
developers (your customers) don't always (perhaps seldom) have control over
our customers (end users).  When it's one box, I add RAM or buy a faster
chip and don't think twice; but, that's not an option for an end user app.


> Well of course there is the cost,
> if the number of machines involved is small, then the cost is not high,
> especially for recently purchased machines with adequate resources.

The other point (continuing from above) is that the cost can easily be
prohibitive when the machines are not under my control and/or there are many
of them.


> >...If there's
> > a "2k is demonstrably more stable" argument, then I'll happily make
that.
>
> Well, there can be absolutely no doubt about that.

I hope you are not just correct, but VERY correct :)


> There are probably still a lot of Windows 95 machines in service, but
> Microsoft themselves no longer support it.

There are a surprising number of them around here, but, even I don't have
much sympathy for being that far back.  I'm using a 95 box now only because:
(1) a machine quit on me unexpectedly and this box was available; (2) stuck
on 9x because of Pen Windows; (3) too busy to do anything about it; (4) a
few file-attributes related problems with my backup system which needs some
work for NT/etc..  The latter will probably just get rewritten in D5.  With
that said, I see 98 2nd and (especially) ME as being too recently release to
yank the plug soon.


> They will no longer support
> (apart from online support) Windows 98 as of June 30, 2003. Note also that
> 98 enters the "Extended" support phase at the end of June this year, which
> means they downgrade the level of free support they provide (I think).

Understood.  Let them be bad guys, not us.



> > My problem with developing on a more stable platform than my users have
is
> > that they find the bugs instead of me.  I often use nicer machines to do
> > certain kind of programming (no need to fight the box when I'm
refactoring
> > or editing a big view resource), but, I test and do a fair amount of
> > development on junk that approximates what my users have to tolerate.
> >
>
> Smalltalkers in general (including us) probably rely too much on the fact
> that we are running in a live environment, literally treading on our own
> toes sometimes, to test our stuff. We ourselves have certainly done this
in
> the past, and still do (though to a lesser extent). Really though it is no
> substitute for proper testing. I get a much more comfortable feeling when
my
> SUnit tests continue to pass after a major change after a run of a few
> minutes, than I do from days of bashing away with classes that might or
> might not exercise that change.

Very true, BUT, that doesn't tell the story for UI testing.  Also, I do a
lot of stuff with quasi real time and very much real-world communications.
There's no way to unit test that kind of thing enough to be sure about it.
I test, of course, but, there's no substitute for getting reflective info
about what happened when something fails.  Many bugs have shown themselves
once and been squashed as a result.

BTW, that's my other big fear re a MS-based VM.  The crash dump is likely to
go away, and that would remove a very effective tool.


> Yes you have to test in the production
> environment, but it makes sense to develop in the most productive one.

I agree, but only to a point.  But note that I do seek out fast boxes for
many tasks.  Things that manipulate my code are best done with the most CPU
I can get; things that excercise my code are best done on the slowest piece
of junk I can find.  For stripping, I use fast boxes: no point in waiting
longer for the result.  I wish MS would go back to the days of making each
developer have one slow machine.


> Even
> just the time saved from not having to reboot 5x/day raises one's
> productivity a lot :-).

No argument there - you should see this thing reboot :(


> A specific problem on 9x is that the resource
> limitations are just too tight for a development environment like Dolphin,
> and one is constantly living on the edge.

But so might be user, and the only way to know that is to live with the
hassles, at least some of the time.


> In summary: We haven't made a specific decision not to support 9x in the
> next version of Dolphin. We haven't necessarily identified any particular
> technologies that would prevent us running on 9x in the future, but at
this
> stage we don't want to constrain future products purely for that reason.
At
> some point we will have to make a decision not to support 9x in a new
> release. It will sure as hell be earlier than five years away (Microsoft
> will probably drop support before then anyway,

No question about it.  Of course, I can point to a Win3.1 box that is in
active use.  9x will be in use long after MS drops support.


>  just as they no longer
> support Windows 95), and could well be in the next major release. When we
do
> that we will continue to support the last 9x compatible release, within
> reason.

That'll work.

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Ted

You wrote in message news:[hidden email]... used
> 'home development' environment. ...

When we first set up OA we had high hopes of reaching that market with
Dolphin, if not with the first version then with subsequent versions that we
hoped to make more and more approachable. We now doubt that market actually
exists any more, at least in any size, and conversations with others would
bear this out. This is a great shame, especially for Andy who has fond
memories of the hobbiest home electronics market he was part of as a child.
There used to be a number of hobbiest electronic magazines did feature
articles based around projects to build this or that (so he tells me, I'm
too young to remember :-)). Early PC magazines had a lot of articles on
programming, but it is usually relegated to a small piece at the back now,
if anything at all. If the hobbiest market does still exist, then it seems
likely that it has largely moved to "free" software such as Squeak.

>...People that want to try Smalltalk typically
> 'play' with it at home for a while before trying to convince their
> management that they've found an environment with serious productivity
gain
> (right Bill?). And of course they pick up Dolphin as that is the easiest
to
> use Smalltalk version around.

Perhaps, but it seems likely that the majority of those individuals would
also be those likely to upgrade the OS on their machines within the next
year or so.

> ...
> ME falls in the 9x line, doesn't it? So basically all people that bought a
> pc for home use up to autumn 2001. People typically replace their pc after
> 5? years, ...

5 years ago Intel had only just released the Pentium MMX running at 166Mhz
and 200Mhz. These machines are only just about useable today (I know, I use
one for testing Dolphin with Windows 98). The useful lifetime of a PC is
probably about 3 years.

>...so when D6 comes out in 2003, then 3/5 of the home users will have
> a 9x machine....

I seriously doubt that. It will be interesting to monitor the figures (does
anyone know of any good places to find stats on the size of an OSs installed
base?).

>....(they can of course just buy my book and use the free D4VE
> that comes with it ;-))

Absolutely, or they can use D5.

Regards

Blair

"Blair McGlashan" <[hidden email]> wrote in message
news:[hidden email]...
> I seriously doubt that. It will be interesting to monitor the figures
(does
> anyone know of any good places to find stats on the size of an OSs
installed
> base?).

I guess if you are using hitbox or similar web statistics you might be able
to guestimate from that data. Here is the data of www.landjob.com  (the
first one I have found with public stats) for January:

            1.   Windows 98    18,816    40.47

            2.   Windows 2000    9,677    20.82

            3.   Windows ME    5,982    12.87

            4.   Windows NT    4,350    9.36

            5.   Windows 95    4,063    8.74

            6.   Other    1,993    4.29

            7.   Macintosh    1,430    3.08

            8.   Windows 3.x    80    0.17

            9.   Linux    32    0.07

            10.   WebTV    25    0.05

            11.   Sun SunOS    23    0.05

            12.   IBM OS/2    7    0.02

            13.   Windows CE    5    0.01

            14.   FreeBSD    2    0.00

            15.   Hewlett Packard HP-UX    1    0.00

            16.   IBM AIX    1    0.00

            17.   SCO Unix    1    0.00

                Total    46,488   46,488   100.00%



rush



begin 666 blank.gif
M1TE&.#EA`0`!`)'_`/___P```,# P ```"'Y! $```(`+ `````!``$```("
&5 $`.PH*
`
end