Journal jd's Journal: Software announcements (or: how to irritate JD) 3
This rant has to do with software announcements. I covered this to a degree in a previous journal entry on the secrecy of some open source projects. This time, I will be more concerned with neglect (the known version is truly ancient, compared to the published version), quality (compare the Slashdot description with the Freshmeat one for the same piece of software) and reaction.
Neglect is a big one. I own 113 project records on Freshmeat and have bookmarked an additional 303. Why so many? 303 bookmarks is a lot - can't I just look to see when the project updates are announced? I would, if they ever were. The bookmarks are reminders of correctly-assigned records that the author can't be bothered to maintain. If they get updated at all, it's because I updated them. With the sheer volume of projects involved, you're damn right if I sometimes think some of the bigger Open Source consortia that develop these packages should be paying me for my time. Globus is no small concern, it's a friggin' international collaboration of multinational corporations. Why are they depending on volunteers to take on the unpaid, thankless, tedious task of fixing their neglect?
Ok, what about those 113? How many do you think I actually created? I'd say maybe half, the others I picked up usually because the owner no longer existed. In a few cases, the records were so stale and decayed that the last update predated the owner field, yet the software has been continuing just fine. Again, that's not good. At best, it means that inaccuracies or other reports will fail - nobody to send to.
Freshmeat is not the only software inventory out there, although it's the only one I make any effort to assist. I've assisted a few paid sites as a consultant, and frankly the stagnation there was even worse. It would be so easy to spend every waking moment just bringing these databases up to speed. We're talking extreme neglect not in the hundreds of records but in the tens of thousands. These are professional sites, paid by customers who want accurate information. They aren't getting it. What they get is something that could be anywhere from a few days to a few years behind reality. Frankly, those customers would be infinitely better off buying a giant disk array and using Harvest to index every site that Google reports has at least one page with the word "download" on it. It would work out cheaper very quickly, and you can be sure of how fresh the information is.
Ok, what about quality? If there's no freshness, then quality is automatically suspect, as projects are evolving entities. They're not fixed for all time, except in rare cases. Ignoring that, though, how accurate are announcements as a rule? Not very. The quality of information is generally fairly poor - sometimes because the person providing it doesn't really understand what is being communicated ("Chinese Whispers") and sometimes because the information simply doesn't exist and has to be inferred from the meager clues that have been left. Sherlock Holmes may be a great detective, but he is also a work of fiction. And if anyone did have those skills, do you think they'd be spending them on correcting project records? Where it's good, it can be truly excellent, but since it would also take someone of the power of Holmes to tell you when the information is good, it's not that useful. If the only way to tell is if you already know the answer, you have no need to be able to tell.
What about reaction? Well, let me put it this way. Atlas' official version is 3.7.29. Fedora Core 7 beta 1 uses version 3.6.0. The official version of Geant is 8.2 patch-level 1, but Fedora Core 7 beta 1 uses version 3.21. I've done some experiments with my own Open Source projects and have found that updates and patches follow the laws of Brownian motion. It is simply not possible to predict if/when/how updates will ever occur within a single distribution, but across all variations of all distributions, the net rate of pickup and refinement is more-or-less constant. This is, of course, completely useless to most users - even those with subatomic vector plotters.
Overall, it's a nightmare to find what you want, a bigger nightmare to determine if it is actually what you want, and a total and utter diabolical nightmare from the 666th plane of hell to determine if what is actually available in any way reflects what it was that you thought you were getting.
Well, okay, but... (Score:2)
Nice suggestion. (Score:2)
It should be easy to produce a meta-record that everything from Freshmeat to Tucows could scan and use directly, with a specific focus on making it machine-readable for such databases. Y'know, that
Re: (Score:1)
When corporations use open-source products instead of rolling-their own, that's fine, but when they pay their developers to modify or correct problems in the original software, and then don't help everyone else by redistributing it, that's just a plain old problem to me.
But in more general terms, I agree with your sentiments, and feel the pain of the plight. I also commend and applaud you sir on