Subversion v. Perforce. Collabnet replies
Hooks and scales
Having allowed Perforce to comment on Tim’s review of SCM tools, we have to extend the courtesy to CollabNet, his other victim, as well.
Bas Nijjer, director of UK sales at CollabNet, points out that “the BerkleyDB database wedging problem noted by Tim has been addressed by Sleepycat and the enhancement will come with the next release of Subversion”. He also claims that “enterprises are turning to Subversion due to its focus on web deployment from design through implementation”. Well, the ones who like web deployment probably are; and Subversion's open source roots are probably attractive in some circles too.
He is “concerned that the proximity of the discussion on offline operations and the discussion on hook scripts may lead readers to think that hooks are client side operations when they are, in fact, server side operations”. Apologies to anyone who got the wrong impression.
Bas concedes that “Subversion's integration with Microsoft's Visual Studio may be a bit weaker than anyone would like” but, in extenuation, points out its “strong integrations with other IDEs like Eclipse [and] pending integrations with NetBeans and Oracle's JDeveloper”.
Tim didn't cover scalability in his review - it's a little hard to test realistically in the context of a short review. But Bas is very proud of Subversion’s scalability, which is “a testament to the diligent efforts put into a product that is still in its 1.x release stream”. On the other hand, Perforce doesn’t exactly run out of steam on enormous projects either.
Bas noted one omission: WebDAV “which can help the non-developer project member to use version control without having to have a separate client or to even have to consider executing version control”. I think it was mentioned, in fact, but only in passing. You can’t cover everything in detail in a short review – and we think that if we write a book only a few readers will plough through it. We can always extend discussion in here! ®
In praise of being incomplete
"There is no such thing as a short review, there is a complete or incomplete review". Yes there is, we published a short review. It was also, of necessity, incomplete.
I don't think computers are as exact as you think they are. Many have trouble representing "one and a third" accurately and testing 100% of the logic paths through a program is usually infeasible. We live with it.
In ten years in technology journalism I've met all combinations of "short", "long", "interesting", "entertaining", "boring" and "turgid" in reviews - but I don't think I've met "complete". And when I was working in IT it never occured to me that a published review could be more than part of the input to product selection anyway.
I asked our reviewer, in effect, to explore two "fit for purpose" products and to discover what purposes they were fit for (although not in those exact terms). I think he did that and if the results attract you to one or other product, you are equipped to do your own "due diligence" in detail, with regard to your requirements.
We're not reviewing consumer toys here. I've been on Perforce training course and have 300 page books just on Perforce - you could do the same for Collabnet. I don't believe that a final paragraph listing the "features" (whatever they are) that we didn't review in detail would achieve very much (we'd just be arguing about the cut-off for the list). But you do have an opportunity, here, to bring these points up - and Collabnet and Perforce can comment on omissions that they think matter.
Completeness of reviews
Many thanks for your feedback and taking the trouble to comment.
I should think on that basis all reviews are incomplete. Even the comparison charts you mention, while they can be helpful, are flawed insofar as they reduce each feature to a tick or a cross, disguising important differences of implementation.
Subversion is free and Perforce is free for up to 2 users, and I'd encourage you to try them out if you have unanswered questions.
Limitatoins of a short review.
It appears from the Collabnet feedback, that El Reg reviewed Subversion against Perforce, and omitted key features of Subversion (and possibly Perforce).
In each instance, El Reg states, "can't cover that in a short review".
Computers are an exact science. There is no such thing as a short review, there is a complete or incomplete review.
What El Reg is really stating is that to maintain the length of text as short and readable as possible, to provide a real world usable article, they published an incomplete review.
This is extremely lame. There are more ways to protray feature comparison information in a review than to describe it textually, eg. A feature matrix - a graphical representation which can be reviewed in 5 secs.
At the very least, if El Reg are going to publish an incomplete article for the sake of readable text reduction, they should include 1 final paragraph which mentions the features that they didn't review in each product for the sake of keeping the article readable.
A review is either complete, or incomplete. El reg seriously failed here.