Where is CST Going?
We have a roadmap published on the Wiki, but I'd like to take a look at one of the areas that we are focusing on: ease of deployment. What I'd like to do is ask our community to rank a few things:
- Ability to ship a Cincom Smalltalk application as a DLL/Shared lib
- Easier integration of C libraries
- Easier creation of a Smalltalk runtime image/executable
Of those three things, which one most needs work? I have my view of the relative importance of them, but I'd really like to know what the Smalltalk developer community thinks. Either add a comment, or drop me a line.
Comments
[Tom Sattler] September 14, 2006 9:55:06.066
Put me down for B), easier integration of C libraries.
I have become frustrated with C header files that simply will not parse properly.
[] September 14, 2006 10:13:40.820
All of the above :-) If I had to pick one, it would be c). It would be nice to create a 2k "hello world". I know that's easier said than done with ST, and I know you guys are working on it.
[Robin Barendregt] September 14, 2006 11:17:13.959
I'll take B in the absence of an option to finish the plugin primitives port.
Re: Where is CST Going?
[ anonymous] September 14, 2006 12:06:35.220
Comment by anonymous
C Lib Integration
then dll/shared libs
then Runtime exe
option C, with some tweaks
[ Troy Brumley] September 14, 2006 12:53:58.018
Comment by Troy Brumley
Can an image be "segmented" or built from disparate files?
Here's what I'm thinking of. In the .NET world, we have all the framework stored in a directory (actually multiple versions in multiple parallel directories, and I can switch easily). I would like to have the ability to have a similar run time framework for Smalltalk, where the VM and base image and parcels are stored in a central location. This would allow me to ship that very small (2K not possible, but 64K might be possible) executable that another commentator mentioned.
As a rough first cut, the shipping run time image would be the results of a perm save as and not modifiable, but my application image or parcels would be merged with it at run time somehow.
option c: runtime deployment
[Alan Wostenberg] September 14, 2006 13:14:32.062
I vote for C: easier runtime deployment. Pass the joel test item#2 http://www.joelonsoftware.com/articles/fog0000000043.html First, make it a one button operation to run the SUnit tests and build the image. Second, make it do that automatically by detecting Store changes (similar to cruise control) in a dedicated build machine.
[Hiren] September 14, 2006 15:13:43.828
[Joerg] September 14, 2006 16:51:01.131
My firt instinct was to say " Easier creation of a Smalltalk runtime image/executable". But then when I reflect on the process in our shop we've solved that problem. I know it is constantly raised as a issue, but for us it really isn't, we worked it out, added some automation and tried to forget about the experiance. I suspect the same will be true for most VW development teams.
Of the other two we have a current need for a smallish DLL that can easily linked into programs written in other languages. We also have an ongoing need to integrate libraries written in other languages, primarly C. So my vote is for "Easier integration of C libraries" by a hair.
Re: Where is CST Going?
[ Michael Lucas-Smith] September 14, 2006 20:48:35.741
Comment by Michael Lucas-Smith
Most know my preference is for option B, followed by C and then A.
A > B > C
[Christian Haider] September 15, 2006 2:51:50.914
You are asking to compare apples with oranges... A is a new feature, while B and C is perfecting existing ones.
Like Joerg, C is basically solved for me (idiosyncratic and fragile (wr. to updates), but working).
B would be nice, but I dont use it much. For me, COM would be MUCH more important - why is nobody discussing VW as a VB replacement? - everybody seems to be focused on Java and servers...
A would be soooo cool. I'd love to make a smallCharts plugin for Excel, Illustrator, Bloomberg etc. This feature would expand the attractivness of VW dramatically (in my view). This would be even cooler, if the DLLs could be small.
do all
[Christian Haider] September 15, 2006 2:58:46.747
sorry, my last comment was misleading...
I think, all items are important and I hope Cincom will continue to work on all of them.
B -> A -> C
[Denis Johnson] September 17, 2006 3:01:11.828
Our vote is B then A then C
We have solved C in-house, although the RTP route is a pain at times and somewhat brittle, once done we rarely need to deploy a new image even though we have mutliple products out in deployment all on the same base image and common parcels. For us the most important is B which inlcludes much needed seamless integration with C++ along the lines of Ian Upright's SWIG Smalltalk. Finally followed by A.