Comments on: Full Speed Ahead?
We've got an interesting situation. Normally, when I release a developer test distribution, it's fairly well tested. It generally works... that is, I spend a fair amount of time on unit testing to make sure that the dist is worthwhile.
However, that's about to change. The conclusion of recent discussions with key developers indicates that they would opt for more frequent releases over better tested releases. In other words, move the major burden of testing from me to developers.
I've pointed out the problems with this approach, and maybe I'm just old fashioned, but it generally goes against my experience and development practices.
For example, with regard to the current R3.A100+ releases involving the extension API, if we just "dump the code" out to the dev-comm, we have:
- Parts of the API may not work at all, because of course, they are not tested. There will be some really serious errors.
- Parts that work incorrectly are just going to confuse developers, and they'll get frustrated.
- Basic functional testing will just flood CureCode with tickets, most of which should not even be posted there.
- Without an official testing approach/policy, there will be no consistent test suite over time... just a bunch of hacked together test fragments.
Robert and others assure me that these issues are either not a big problem or can be solved.
Well, I'm fine with giving this a try, but it also means that other members of the dev comm will need to step up and deal with the resulting chaos and anarchy... turning the noise into information, and relaying it sanely back to me.
So, this is new. We've not tried it before, but I'm willing to do so. I'll let Robert handle the management and organizational aspects of this process, and we'll see where it leads.
To me this is an experiment. It makes many fundamental assumptions about motivations, efforts, and organization. But, I think we'll learn something, one way or the other. If it works out, then it is of benefit to all primary developers. If it does not work out, then we'll just go back to how we've done it for so long.
Are you ready? If so, then "damn the torpedoes, it's full speed ahead."
Build of a101 would not go through on Cygwin for ole Win32 XP but with some ln -s things went fine under Msys
First MSys build failed when could not find 'copy' but fine on second pass with link to 'cp'. Will move over to EeeBuntu 3 now ...
btw, the a101 r3 built under XP Msys runs fine under Cygwin with rlwrap (which uses some cyg DLL's ) for a command line (as using an arrow key proves to be not good for the next cmd entered after key that fails to get a cmd from history); command line behav seems good in vanilla W32 cmd.exe session with arrow up retrieving cmd from history.
unable to build yet on EeeBuntu ( I build a lot of open-source on this netbook )
RobertS, A101 currently is a Win32-only release.
The approach could be similar to having stable and unstable versions. Make sure to differentiate them well one way or the other so people know what they get.
Don't do it. Test them as much as you can first.
"Without an official testing approach/policy, there will be no consistent test suite over time... just a bunch of hacked together test fragments."
i'm agree and that's what i try to speak with the community those past days/weeks. actually the R3 GUI effort is based on an effort based upon some money raised without knowing clearly what is the task due for that money. And for the future what will be the outcomes of having outside people taking in ostage the project because they claim loans for the project to evolve.
Having the need to ressort to the king dollar as ultimate mean of presure to put at work our lazy comunity is not a good sign sent to the world... and it's not a setup for futur, for a proper design.
when i try to get informations or a serious discution about those aspect all I get are insults. the Gurus are being arrogant and overtaking the project...(by the way i didn't knew robert Münch could lead anything regarding rebol's VID... so if the architect don't know how to build a house how will he ensure us that it's crew won't build the house upside down)
Those gurus are liying to us by saying this way is the only affortable way to make R3 GUI project go on.
I like the fact VID is externalised in hostkit but instead of being the base of a new mature motion in our dieing community on educational level and being inspired to build the new vid in commun based on those 10 past years of experience thank to those undelicate avid of financial profit autoproclaimed Gurus it will be the swarn song for rebol.
Post a Comment:
You can post a comment here. Keep it on-topic.