Smalltalk › Cincom › VisualWorks

Long live all VW apps

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

25 messages Options

Niall Ross

Re: GUI development time share

Dear Joachim,

> I would say Niall's estimation is pretty realistic. ... I'd probably
> have chosen 30+ percent instead of forty, but Niall's calculation
> would still be valid and bring up impressive figures.

I can agree with 30%; I deliberately went a little high rather than
appear to bias the calculation to my preferred view (but conversely, as
you may have noticed, there's an assumption embedded in my figures that
re-doing is only half as hard as doing since, as Steve remarked in his
reply, a significant part of the UI effort is the thinking behind the
code, which a new GUI framework does not change; I should have said
that explicitly. And of course it is all very simplified, just done to
assess the pros and cons).

The converter must work dynamically as well as statically to handle some
of the weird things people (myself included) sometimes do to work
with/around Wrapper. It must analyse and convert, partially evaluating
and partially rewriting, fragments of old UI code into fragments of new
UI code, and it must assemble the result into Widgetry code that an
application's maintainers will find readable and evolvable. AND, which
is my point in these posts, it must succeed in this to a high percentage
before 'type 2b' customers can truly get value from it. (Christian's
WidgetryWrapper would allow us to encapsulate dire cases for a time;
that may move the percentages somewhat but not change the fundamental
point.) As noted, from my own rewriting experience I believe this is
quite feasible, though not easy.

This completeness requirement - the fact that individual increments will
not find eager users - gives the project a technical dynamic that is all
wrong for a community project, a Camp Smalltalk or similar. Hence my
first post's mention of 'funded commercial project'. It may also, and
understandably, worry any possible funders.

I hope this clarifies what I wrote and why.

BTW a customer type not mentioned in Steve's analysis is those VSE
customers thinking of porting to VW. Widgetry would greatly simplify
their UI ports. Its loss correspondingly complicates their plans. (And
they may well not be frequent readers of VWNC. :-)

>> I think I've said all I usefully can so will (try to :-))
>> restrain myself from posting on and on about this

I shall try harder :-).

Yours faithfully
Niall Ross

Cesar Rabak

Re: Long live all VW apps [Just one question]

In reply to this post by Maarten Mostert-2

Maarten Mostert escreveu:

> For those of you who never had the chance to see how Widgetry
> applications actually look like you can check out some views on my
> simple sparetime vwnc application.
>
> Behind the scenes its all Glorp but some of you probably guessed that
> allready (thanks Alan).
>
> *Just one question:*
> Imagine you have gathered all this decisions makers together for your
> demonstartion and then you open up your image (Some of you who were in
> Sales might remember that thrill). Then tell me honestly which project
> you would prefer to be on.
>

Maarten,

I renamed both images to two random generated temp names (to avoid any
chance of giving a hint on hierarchy) and showed them to folks here (not
a crowd but six different persons) and the outcome was 50% - 50%...

Just seeing that images there is no compelling reason to go either one.

my .01999...

--
Cesar Rabak
GNU/Linux User 52247.
Get counted: http://counter.li.org/

Maarten Mostert-2

Re: Long live all VW apps [Just one question]

Cesar Rabak a écrit :
> Just seeing that images there is no compelling reason to go either one.
Thanks Steve Eliot and Cesar to have given the feedback.

Actually this where really as is images, I hope you have noted the
better bitmaps in the tree-view border seporators, decorations and grid.
The ganttview obviously remains to be improved but the difference
between the two views was mainly the different zoom level and some
connections.
I'll try to come back on this later also to show you the Mac views, for
the moment I am seasiding. (A not yet destroyed thing ;-) ).

Rgrds,

@Maarten,

Lynn-2

Re: Long live all VW apps

In reply to this post by Steven Kelly

Well we fit into category 2 but have choosen to build our newest release of our software with Widgetry. We're continuing on this course. One point I haven't seen yet deals with the concept that it may be possible to harm the future to protect the past.

Just how much support and cost Cincom puts into Wrapper in the recent past was not clear to me but seemed very low compared to their other initatives. I think those who have lots invested in their apps (as we do) and use Wrapper can and should keep using it. Seems to be a no brainer - Cincom hasn't been pushing it forward anyway but you seem to lose nothing. Your apps already work and people like them - thats great for you. I'm curious how much input you gave Cincom during the past 6 years as you watched the Widgetry initative develop and if you were putting forward the argument to them that we're paying customers and we want to stay with the old, we don't want the new.

We on the other hand think our new GUI's will be much better based on Widgetry. I imagine folks can mix and match if they want, old with Wrapper, new with Widgetry. Our experience is that Widgetry is well designed and implemented and is easy to work with. What we could never understand is why it was a one man show during the past 6 years. I can't understand the business logic behind this decision at all.

Anyway, I don't know why putting a GUI builder together for Widgetry at this point really isn't a good idea. I'd want to capture the cost of the last six years and get a return on that investment, keep a certain clientel happy who has been patiently waiting for it while not dropping Wrapper from its supported status. Breathing unknow life into it now after so many years of ignoring it along with its known architecture issues still boggles me. I guess I'll just have to wait another 6 years to see what they have in mind. Hum... wonder if I can stay loyal another 6 years. Hope it see's the light of day.

I'd love to contribute to a bounty to get some creative Smalltalkers to build us a Widgetry GUI builder. Maybe there are enough of us out there to make this happen. I don't want to hurt your use of Wrapper going forward but I likewise would like to also move forward with Widgetry. Afterall don't we all seperate our domain expertise from our views and GUI's. Lynn Hales

Steven Kelly wrote

From: Maarten Mostert [mailto:maarten.mostert@wanadoo.fr]
The fact is that except... for your HackerTalkers not a single developer who tried Widgetry came back stating it is not worth the change.

Whether something is "worth the change" depends on more than just "is Widgetry better than Wrapper". Until there is an automated conversion or interpretation of Wrapper windowSpecs and ApplicationModels to make them work in Widgetry, I don't even need to try Widgetry to tell you it's not going to be worth it for me to migrate. Our application has been built in VisualWorks since 1992, with something between 50 and 100 man-years of effort. It doesn't take much maths to show that even to just move with a "word by word" translation to Widgetry would take rather a long time, and give very little real benefit instantly (update speed is not much different, maybe some less flickering). To take advantage of any new features Widgetry offers would take even longer, and still not significantly enhance our application from the point of view of its users. Our users definitely don't buy our product just to look at whether the emulation of a button is pixel perfect, but to use our features. We'd be much better off spending our time on those features than on a massive GUI migration project to stand still.

For the remainder of this message, I'll assume that Widgetry is significantly better than Wrapper - mostly the important thing is that it is internally better, lacking the crusty agglomeration of hacks that Wrapper has become in many places.

Put simply, the less existing investment in VW that people had, the more likely they were to invest in building apps with Widgetry. But generally speaking, the people who actually pay Cincom are the ones who already have a significant investment in VW. By not providing a clear, low-effort, low-risk migration path to Widgetry, Cincom ruled out that migration for most people who pay its bills.

So we have three groups that are interesting:
1) Cincom
2) long term customers who pay Cincom's bills and have large amounts of Wrapper GUI but no Widgetry experience,
3) new customers who pay Cincom little yet, may not have used Wrapper much and have invested in using Widgetry

Group 2 would be happy for Wrapper to continue and be improved, yet recognise that at some point it would be great to have a new GUI framework. They're used to having a graphical GUI builder for one-off UIs, but also build GUIs programmatically when necessary. They've put up with an awful lot of trouble from Widgetry over the last 6 years, since it has stopped sorely needed work on their existing GUIs, and has stretched its deadline to four times what was promised. Group 2 are those that would directly benefit the most from the promised automated conversion, yet neither they nor group 3 are really expert enough in both sides of the conversion to build it.

Group 3 would be happy for Wrapper to be killed right now, preferably with a stake through its heart (thank you Maarten for making that vehemence clear! :-> ). They have no direct desire for conversion, or for a graphical GUI builder (since they have been prepared to build everything by code, they are probably in the 20% of the populace that prefers text to visual). Indeed, they may even oppose work on those areas, as they would divert resources from other improvements to Widgetry.

That leaves us with Cincom. They have the experience on both sides of the conversion. They can't afford to support two frameworks in the long run. They thus have to guess whether group 2 is going to continue providing income in the long term with Wrapper-based applications, or group 3 is going to grow and provide a similar amount of income with its Widgetry-based applications. My guess is that group 2 splits into 2a, whose apps are not actively pushed forward and will not update to Widgetry under any circumstances, and 2b, whose apps are actively developed and would be updated (grudgingly!) given a sufficiently good automated conversion. Long term, 2b will tend to provide income and 2a will tend to die out - regardless of our choices here.

For Widgetry to survive it must become the single GUI framework in VW, in about the same timeframe as VW replaced ObjectWorks, or events replaced polling. For that to happen, groups 2b and 3 both need to push Cincom to provide automatic conversion.

Effectively, Cincom have built a new "house" whilst allowing the old house to fall into disrepair: unless the new house is seriously flawed, it doesn't make sense to abandon the new house and start patching up the old house, just because we suddenly notice we haven't yet hired a removal van to help us move house.

Cheers,
Steve

Maarten Mostert-2

Re: Long live all VW apps [Just one question]

In reply to this post by Maarten Mostert-2

Hildebrant,

Both images look fine to me. Actually, I've been looking for a smalltalk Gantt chart widget. How can one obtain one?

Actually I build this Gantt with my small fingers. It is currently not written as an API and destinated for a multi purpose project management tool I work on.

Rgrds,

@+Maarten,