SCM for AETF

Posted by Admin on 22-Apr-2011 08:12

Following Peter's suggestion, I'm branching off the SCM topic.

I must say, I'm much more in favour of a centralized SCM strategy, preferably a SVN repository hosted by PSC (on EC2?).The original version of the code would be available there and every PSDN user could have a personal repository or a personal branch. Similar to the Perforce Public Depot (http://public.perforce.com/wiki/Welcome_to_the_Public_Depot) that hosts a lot of free available tools from the vendor and community projects. It simply creates more visiblity than distributed GIT repositories here and there. And as of today there is still much more tool support for SVN than for GIT (if I'm not mistaking there is still no Eclipse plugin for GIT).

Owners of branches (PSC for the origianl AETF, customers for customization, enhancements or our own projects)  could decide which features of the others branches they want to integrate into their code(base).

PSDN would be the forum to create visibility about what's there.

That would also dramatically improve the code-share which is really not where it should be.

All Replies

Posted by jankeir on 22-Apr-2011 09:08

Mike has a few good points for SVN. There is always hgsubversionfor people who absolutely want a local repository. Eclipse does have GIT (eGit) and mercurial plugins though, they are just not as polished as the SVN tools.

Posted by Admin on 22-Apr-2011 09:41

eGit does failed me a few times, I'm sure this was my fault of not having use it the right way ... on the other hand the mercurial plugin is great, it has mylyn integration and you can do what ever you want with it.

distributed SCM are best for peoples working remote, not always having access to centralized repository with somehow longer periods between commits... just like about all open-source projects

it might sound more complicated that it really is, just go make yourself an user on bitbucket and start playing with it a bit... once you get started you'll love it (or not, I guess there can be only two options).

https://bitbucket.org/

Posted by stevenseagal008 on 22-Apr-2011 10:41

I have no ties to this project and don't plan to, but I do have lots of experience with SVN and git both so I will add my opinion.  I've switched to git and will never go back to SVN.

git and Mercurial destroy SVN when it comes to distributed development, because that's what they're written for.  You can still do a centralized model in these SCM's quite easily by agreeing to use a specific repository as the "blessed" repository.

Continuing to use SVN for distributed development is kludgy, in my opinion.  I hope those who insist on SVN will be volunteering to fix all the merge conflicts that will crop up when merging branches?

Posted by stevenseagal008 on 22-Apr-2011 10:43

Yes, but the core is still SVN and you will still have to deal with merging conflicts when merging branches much more often than with a distributed SCM.  Have fun with that!

Posted by Admin on 22-Apr-2011 11:13

hahaha, don't put it that way... you'll scare them from the start

it  simply depends on how they plan the contributors to submit code changes...  they may go for a controlled environment with some sort of submission  procedure and only one master user that actually commits changes in  repository (read-only access for the rest of us mortals) in that case  anything will do. if there are going to be multiple contributors with  read-write rights working on different parts/modules then I also  strongly advise them to go with a distributed scm, unless of course one  likes to experiment that kind of pain you just mentioned

Posted by Thomas Mercer-Hursh on 22-Apr-2011 14:17

Begins to sound like POSSE ... which is not a good thing.

Posted by Admin on 22-Apr-2011 17:10

Begins to sound like POSSE ... which is not a good thing.

I don't see where you are getting that impression from. Nothing written in this thread has anything in common with POSSENET.

A sample, a starting point for people looking for an idea of a way how a new architecture of their app might look like is not Dynamics. But everybody could look behind the scenes of the current sales demos. It would certainly have an educational value

When Progress then would offer a central place where everybody could either branch off from what's there or start something new or provide existing stuff they already have there is nothing in common to POSSENET where the best you could do was to add a ZIP to Bugzilla.

I believe that people who have no experience with N-Tier or OO wouldn't be able to choose from your library of competing alternative components because they have no clue what they need and no metrics to decide which of the competing alternatives is the right for their security module - and which components play well enough with each other.

But still we could make those alternatives available on a central place.

Believing that Peter and Mike O. are lazy like most of us are, when there would be something really good in your version of OERA or mine, they probably would use it - it not, it would be an TMH or MFE or JLS variant - on the same site, in the same catalog.

Key is, that everyone of us can provide and maintain code in an SCM and make it visible next to PSC's offering. Without going thru a process similar to Apple's AppStore submission process.

Posted by Thomas Mercer-Hursh on 22-Apr-2011 17:37

I think I have been misinterpreted.  An "open source" project in which everyone outside of PSC submits possible changes, but the only ones which actually get incorporated are the ones that PSC decides are good, sounds a lot like POSSE.  Admittedly, handling open source for a whole application is difficult, very difficult.

I think a component library is a lot simpler.

Yes, people need to know how to assemble the components ... but there is also more than one way to do that.  If someone wants to pick up a security component or a logging component or whatever and simply begin grafting it into their legacy, non-OERA application, isn't that still a plus?  Shouldn't the documentation cover any dependencies ... and, with good design, shouldn't the dependencies by minimal?   Shouldn't one be able to provide one's own implementation of any related services, e.g., if one's need or context is different.

How exactly does one show off that there is more than one way to do OERA?  Only if there are multiple reference implementations.

Posted by Admin on 22-Apr-2011 17:55

How exactly does one show off that there is more than one way to do OERA? Only if there are multiple reference implementations.

Isn't that exactly, what I wrote?

One reference implementation from PSC, one from you, one from me and one from everybody who has something to share?

You decide what get's in yours, I decide what get's into mine and PSC decides what gets into theirs... All available next to each other with documentation and a catalog that allows people to compare.

Posted by Thomas Mercer-Hursh on 22-Apr-2011 18:04

How likely is it that any of us other than PSC are going to create a reference implemenation?  Component, yes, but not a whole implementation.

I just see the purpose as being very different.  AETF is something where one wants to download the whole thing, play with it, crawl around in its innerds, think about its idea and whether one agrees with them.  The catalog is something where one is going to go looking for a particular component type and see what is there and download just that one or two selections for further evaluation.

Posted by Admin on 23-Apr-2011 00:20

How likely is it that any of us other than PSC are going to create a reference implemenation? Component, yes, but not a whole implementation.

It's not unlikely that you are going to be right here. But I've always written of a place where everybody could share whatever they have. That could be a mini specialized component.

It's not likely that anything good gets created when 100 different coders or architect without any kind of steering create a useful or even better AutoEdge. So it must be Progress decision what they branch into their reference implementation. Anybody else or a group of individuals could create an alternative RI.

It's important that it is open for all kind of projects, like the Perforce Public Depot. And I believe that it should be driven by the vendor, because I believe that that allows to reach more people.

But what exact alternative for contribution do you suggest?

Posted by Admin on 23-Apr-2011 01:18

mind you that posse was not as bad as it sound... it had scm, forums and  I've even managed to get ganimede as a separate project on it having  full control over the project pretty much as it is on sourceforge.net,  the bitter taste was the way everything ended... not even an email to  say we should clean-up because the bar is closing

so, the expectancies are quite different here... collaboration over  show-case, if I remember correctly none of those did work for posse,  code submition and comunity involvment was very limmited and having the  project on PSC platform didn't make a big difference either.

bottom line, I have no idea what those guys are hopping to get out of it  and what their intentions are... so far it's like some code that they  put toghether for a show-case application and they got the green light  to release it publiclly (as-is), if there is going to be anything more  after that I guess we'll have to wait and see.

This thread is closed