This was originally posted on Facebook as a Note on May 12, 2008. More administrivia at the end.
Google Reader (Google’s [now-defunct] online RSS feed reader) is on the whole an excellent application (as much as a “web application” in HTTP can be … which is a subject for another Note). However, it does not scale well.
I am subscribed to 2140 feeds. Most of these are news, release, documentation release, or wiki change feeds for Sourceforge projects that I might be able to use or borrow code from in my own strategy game, Strategic Primer. Most of those are empty, since most of the projects either didn’t release anything, didn’t have news, or didn’t have documentation. (Wiki update feeds were only available for projects with wikis.) Unfortunately, adding a single feed, or importing an OPML file full of feeds, made my browser freeze for several minutes. (I will admit that I have not tried this since an apparent speedup that mitigated somewhat but did not remove the below concerns.)
A second scalability concern is unread items. For a long time — because I also added several feeds of political columns and let them go unread for some weeks as I tried to catch up on the Sourceforge feeds — I had approaching, and sometimes well over, a thousand unread items. (After that limit it just shows “1000+”.) Until I got that down under 50 or so (but perhaps 50 per feed would have worked just as well), it was as likely to tell me I had no unread items — network timeouts on the AJAX, I presume — as it was to give me an accurate count.
Yet a third issue is reaching the bottom of a list of items. In my case, it’s my starred items. I’m converting my bookmarks into del.icio.us bookmarks, and for me “starring” means “go bookmark this later.” So I wanted to bookmark things, the earlier ones first. But there’s a slight problem: I have over ten thousand three hundred starred items (dating back to December), and it tells you “over 20 items.” I go to the end of what it will show me, and it increments by 40. Repeat. Over two hundred fifty times. And if, for some reason, I (accidentally) leave or refresh that page, I have to do the whole thing again. There is no way — as there is with every feed — to get the items in reverse order (though that only gives you the past 30 days in any case). No way to just export the list as a bare HTML file with all of them (short of figuring out their protocol).
[Ed. note, as of 2009, before Google scrapped Reader: I have since removed most of the Sourceforge feeds–which I could only do in a batch by using an external library in a program I wrote myself–and the delay on adding a feed no longer makes the application freeze, but the application still takes at least fifteen seconds to do anything with several hundred feeds and over a thousand unread items. And don’t get me started on the idiocy of making items over a month old go away when I’m trying to work my way through them all. Unfortunately, I have been unable to find any good alternative.]