FeedTree coverage.
FeedTree had a pretty good week. After I released version 0.7.0, I sent the article to Slashdot, where it was run on the front page. The comments told me a few things:
- This is not a hot-button topic for Slashdotters; there were very few comments at all. (Furthermore, the Slashdot Effect was nowhere to be seen on our servers; the big iron that runs the main site didn’t really notice, and even the Pentium-III that runs the Trac server and Subversion repository had no issues.)
- A few people don’t get it. “Isn’t this just NNTP?” “Why not use BitTorrent?” etc.
- A lot of Slashdotters (those reading the article, anyway) do get it, and were quick to set straight the naysayers and the clueless (e.g. 1 2).
So after a front-page Slashdot article (which generated approx. 3500 direct hits to feedtree.net), you’d think I’d be drowning in users, right? Well, we spiked at 30, and are currently hovering around 15 users. That’s … well, it’s not a lot. I have a few users who are patiently waiting around while I try to figure out why their routers are blocking Pastry packets, but I think this is pretty much all I’m going to get for a while. Hopefully it’s enough to generate some meaningful data!
Psst, FeedTree users: you do know that if you get more people to use the system, your own service (speed of updates, amount of polling, etc.) should improve, right? Just a suggestion.
Fortunately, the Slashdot story did result in a number of mentions here and there across the Interthing. Perhaps the most surprising was a podcast mention; the GeekNights ‘cast discussed FeedTree in their 2/20 show (starting at 53:46—clip). Thanks, guys!
Other thoughtful, sometimes critical, mentions worth a mention: Brian Dennis, Diwaker Gupta, Netemic, Abhijit Nadgouda, HowForge, among others.
7 responses
[posted here and to feedtree-users]
Allow me a minute to explain why I think your slashdotting was so unproductive, and the factors that, unless addressed, will cause FeedTree to fail.
The people who need to use FeedTree are not the people who are having the problem FeedTree solves.
The FeedTree homepage states two benefits:
To users: Get RSS entries faster than every 30 minutes.
To publishers: Save bandwidth.
To publishers, this is an excellent benefit. However, since the user base is so low, it’s not particularly tempting at this time. Besides, if they go with Feedburner, they get the bandwith benefits plus user stats, advertising revenue, and various other services. Single point of failure? Eh. That’s someone else’s problem.
To users, the benefits are negligible at best. First up, I think that few people care about getting RSS feeds in real time. Most people I’m aware of that use feedreaders are more interested in “aggregation” of feeds. Gathering a web’s worth of stuff they can check in with a couple of times a day. After lunch, or in the evening.
And heck, if you crank the settings on your feedreader, you can get news every minute. Sure it’s “rude”, but it’s not me that’s hurting, it’s the publisher.
So you’ll have a bitch of a time convincing any decent number of users to download an extra program, install it, re-configure their newsreader, and punch the appropriate holes in the firewall.
Or to put it another way: I’m a big fan of the project, I’d love to see you succeed, but I don’t use FeedTree with my reader. Too much overhead and hassle for not enough benefit. If you can’t convince me — a friend with a keen interest in RSS and P2P — to use FeedTree, what hope the rest of the world?
So, little benefit to users, and little benefit to publishers if there are no users. Catch 22 much?
The only way I can see to increase FeedTree usage:
1) Remove the requirement that you have to punch holes in your firewall. Your argument that people punch holes all the time for BitTorrent is flawed because a) People actively want the pr0n/warez/whatever that BitTorrent delivers and b) BitTorrent works (albeit slowly) without the firewall tweaks. The firewall tweaks just improve your service. Unless you make this change, no feedreader is going to bundle FeedTree into their app — the support time and costs would be phenomenal. Which leads to…
2) Get feedreader software authors to bundle FeedTree. People are only going to opt-in to FeedTree if it’s painless and takes no effort beyond hitting “Yes” after not reading a dialog box.
My $0.02. Curious to hear your reactions.
Rod.
comment posted at 8:31 pm on 25 Feb 2006
Another reason you might not have seen much of a spike from Slashdot: dedicated vs. web clients (aka “everyone uses Bloglines”). Slashdot users are considerably more likely than most to try to use Linux as their prime-time desktop OS. How many decent RSS reader apps exist for Linux?
[Crickets...]
Right. So they all use Bloglines, and have no use for FeedTree.
comment posted at 12:12 am on 26 Feb 2006
I’m not sure about that; Linux users have RSSOwl (which is apparently pretty popular), akregator comes with KDE, and, oh, Firefox and Thunderbird too.
I mentioned this on feedtree-users—do you think it would help if I had a pile of public HTTP proxies people could use? Yeah, it throws away a lot of the scalability, trading it instead for cheaper adoption.
comment posted at 12:34 am on 26 Feb 2006
I’m one of the people who learned about feedtree via slashdot. I actually don’t really ‘read slashdot’ anymore, but just scan the headlines via my rss aggregator, and click through the links if anything seems interesting. If slashdot would export the links into their rss feed, i wouldn’t bother ever loading their page, but they seem to have some technology issues. *cough*
I do use linux, and I use the quite capable ‘liferea’ aggregator that’s built with gtk. But i search for a feedtree package in debian, then started reading about ‘underlying technologies’ and pastry vs feedtree, which I really didn’t care about.
Then I saw it was written in _java_, and I lost all interest. Java is just completely unappealing to free software folks, to tasteful software programmers, and most practically Linux users. I decided that if the idea had significant merit, someone woudl eventually write a version in a language that doesn’t use a VM that claims half my core.
comment posted at 3:38 am on 11 Mar 2006
jrodman, I’m sorry you feel that way. I don’t think Java would have been my first choice either, not so much because of free-software politics but for technical reasons: bulky virtual machine runtime (although a 30 MB image, with substantial shared pages among multiple JVMs, doesn’t really seem like that big a deal these days) and language verbosity (anonymous inner classes are the world’s ugliest function pointers, including C!).
That said, the FeedTree idea (which has merit, IMO) relies on a number of other significant inventions that do not currently have an implementation in C++ or Python or whatever. FreePastry is a powerful, complex, mature piece of software that is challenging to reproduce. As a graduate student researcher, I have to carefully balance my time budget: Do I implement the innovative features of FeedTree on top of FreePastry, leveraging the Pastry team’s expertise and prior hard work? Or do I go it alone and reimplement everything? It was sort of a no-brainer for me.
I’ll point out that you don’t need a Debian package to run the client software; just grab the “jar” link from the download page, unzip it somewhere (somewhere stable, not /tmp), and launch the
ftproxygui.jar(either from within your gtk filebrowser, or from a shell withjava -jar). I’d be very interested to hear your feedback based on using the program.comment posted at 9:17 am on 11 Mar 2006
The optimizations in memory use you cite are of course what sun cares about. Making sun java’s memory use lower impact for server side deployments on massive machines.
On a desktop there are typically 0 JVMs running at most times, so sharing memory between JVMs earns no memory savings. Moreover, longrunning java processes on the sun JVM tend to swell to hundreds of megs and manage through some VM magic to keep nearly all of it locked into real memory even when little to no actual software activity is ongoing. The net result is that simple network services written in java occupy half my real ram on a permanent basis, when I’m trying to get work done.
There’s also the issue that there isn’t a debian package for _java_. That is there is no sun jvm debian package because of sun’s obstructive policies. So just installing the substrate your tools require is unpleasant. It’s possible your tools will work with gcj, but my experiences tell me that software I download from the web spit out exceptions about unimpelemented classes.
I’m also quite spoiled. These days most new services and systems are installed with a single command, with possibly minimal configuration. A package that tells me “you can run it from anywhere” from where I sit is not a feature, it is a burden. I want the software to have a good idea where it will live, help me get it there, and keep track of itself. I don’t want to install services into some random directory I have to make up at the time of install and later regret. So a jar file that “you can just run” is actually much more hassle than a traditional package that embeds fixed paths and lands where it wants.
I suppose the .deb would take care of some of these issues, but I just don’t trust it because it’s a .deb that requires a system substrate that Debian does not provide.
Again I don’t think you’re a bad person for using java, or that it is necessarily a bad decision. Obviously it’s cross platform, although I think Bittorrent achieved much more successs than the average java program. So it’s leveraging a sophisticated existing technology, okay. I’m just spelling out the details of why java means “no way” for a not terribly atypical debian Linux user.
comment posted at 6:42 pm on 11 Mar 2006
[...] FeedTree coverage. [...]
comment posted at 5:31 pm on 14 Mar 2006