Hacker News Reader

Comment by rb2k_ | Hacker News Reader

> because IBM's tools & tech could greatly aid Compose in developing their stuff faster, more robustly, etc.

Could you give an example for that?

Seriously? IBM has a whole portfolio of technology with plenty being very competitive. It's one part of how they're the largest tech company. Getting their software free or hardware at cost internally would be a strong enabler for about anything.

This would be a nice start for a division or project: Rational w/ all analysis & productivity addons; the management software; cloud services (i.e. backups); security services/software (basic); an IBM i (comes with DB2) for critical stuff. Add in some servers and storage appliances [from IBM] where needed as growth occurs. Also remember that that Microsoft used to run their whole business on one AS/400 until they were called out and replaced it with 20+ Windows servers to match performance/reliability. Funny stuff.

They have all kinds of good tech. A permissive policy on its use, even if for limited time, could greatly aid internal projects and growth of companies they acquire. If I was them, I'd throw my weight into everything on the software side given it would be as easy as an order, a license, and a download. ;)

There's a lot of examples. Have a look at Bluemix.net, Jazz.net, Bluemix Devops Services (https://hub.jazz.net/) and some of the DevOps tooling like Urbancode.

I assume he's talking about their Rational suite: https://www.ibm.com/software/rational

I actually have an awful lot of time for Rational Team Concert. The VCS is quite well tuned to include some of the notable benefits of DVCSs, while keeping things a bit simpler to work with. Work items are really nicely integrated too.

I'm conflicted on the VCS. It seems powerful, but the "bag of changesets" (instead of linear chain of changesets) model gives me issues when performing operations that I would commonly do on my git repos. In scripts I have written, I've encountered two changesets on the same stream occurring on the same second and accepted them individually in an incorrect order, which put the workspace in a screwy state.

(Also hitting the server for every little operation kills me. Even the command line tool just communicates to a java daemon to go hit the server, arrrrrgh.)

On reflection, I suspect a lot of it is less the actual VCS itself, and more the way it gets exposed in the UI. In contrast to git (where I'm 100% command line), I spend most of my time with RTC using the GUI, which is reasonably well directed towards the most common needs.

I think I also have a lot of positive thoughts about RTC because I remember how easy it felt to get started with it. With git it took me a little while to internalise how it worked, and I screwed up my repositories beyond my newbish ability to recover a few times. When comparing each to SVN, RTC provided me the improvements I cared about most (easy sharing, interim 'commits' prior to pushing to mainline), with very little learning overhead.

Combine that with really excellent work item/defect integration and I think it's a pretty compelling product. As you say, the constant communication with the server is a definite bummer. I think that comes out of the more corporate focus of an IBM product - for a centralised team sitting in the same building as the campus, a bit of server communication isn't such a big deal.

I'd be pretty salty about having to use the VCS if it wasn't so well integrated with the work items. Definite improvement over GitHub in that regard. (I've still entertained the thought of writing some sort of git translation layer, but I don't know either well enough to translate some of the crazier edge cases.)

It is an interesting point, but it really shouldn't matter what order you accept the change sets in as long as you accept them all, correct? Was it something specific to your project that caused there to be issues?

I'll try and reproduce the issue, but I think somehow the workspace knew it was in an inconsistent state, but wouldn't provide me the information to not put it in that state in the first place.

This was "fixed" in later versions of RTC with --accept-missing-changesets [0]

[0]: https://jazz.net/library/article/1372

For the VCS, do you mean Clearcase?

Clearcase was a good idea in 2001, when the alternative was CVS. In 2015, I am wholeheartedly unconvinced it offers anything over git. (And I'm not a devoted booster of git.)

(I was nominally responsible for a Clearcase setup at Ericsson in 2001. I mentioned it in passing on the 2002 version of my CV, and no later version. I still get pings from those last few Clearcase shops, desperate to find someone willing to touch the thing.)

You may disagree, and perhaps Clearcase has substantially changed its model since then. What in detail do you like about Clearcase over (say) git?

Ah, I have no experience at all with ClearCase - I believe the VCS for RTC is a newer product with some conceptual overlap, but I'm given to understand it is typically substantially quicker and easier to get up and running with.

That's good news :-)

The last one I remember was Karger et al's paper on lessons they learned building Caernarvon smartcard OS for EAL7 security. They mentioned in the paper they chose the IBM tools for static analysis, etc because they were free for internal use. So, I'm trying to figure out (a) if that's IBM tools in general, (b) who it applies to, and (c) how much.

If tools were available, I'd certainly consider Rational's modeling, testing, and cross-platform features as worth due consideration. :)

RAD is some of the worst software I ever used, I get angry just thinking about it.

Reply on news.ycombinator.com