RefWorks
A "Personal Web-based Database and Bibliography Creator"

Apparently, people at IU South Bend asked several users for comments about different tools and ended up with RefWorks. Can see why. In terms of ease-of-use, it's very good. And it has many interesting features, including some that aren't found in the typical dedicated desktop applications.

I must admit, I'm rather impressed with their rate of release. They seem to follow the typical open-source model of "release early, release often."
In fact, although it's proprietary/closed-source/commercially-distributed (through CSA) and not necessarily inexpensive/free-as-beer, it's almost open-sourcesque in its approach. At least, much more so than Thomson/ISI products.
Funnily enough, CSA integrates with Endnote (made by ISI) better than ISI products do. 😉

Of course, there are several good bibliography solutions around. A cool open-source one is BibDesk. Originally meant for BibTeX data, it now does much more and serves as a cool solution to autofile PDF versions of academic articles (realising part of the dream of an "iTunes for academic papers"). What's neat about RefWorks is that it can be shared. Not only is it possible to make any number of accounts for specific projects (very cool solution for classes) but it has a specific tool for reference sharing. Didn't use it yet but the rest of the program is good enough that RefShare can't be all bad.

Well, this is getting into a pseudo-review, which would be much more difficult to do. One thing that's rather impressive for an online system is that it accepted a submission of tens of thousands of references from an Endnote file without complaining too much (apart from server delays). So they don't seem to have a limit in the number of references.

Which leads us to an interesting point on reference software. [Start rambling…] A given item, say a reference to a journal article, will be present in many people's reference lists. Most of the data should be standardized for all occurrences of that item: author name, publication date, complete title… Some things are added by the user: date accessed, comments, reading notes… In good database design, RefWorks should only keep one copy of that item (with the standardized information) and have links to that item in people's lists. The customized info could probably be streamlined and will probably not amount to a lot of data. Now, there's an interesting side-effect of this as common references should in fact be standardized. One of the most nonsensical things with online reference databases is that you might have "Smith-Black, John D.," "Smith, J.," "John Daniel Black-Smith," and "Black, J.D.S." referring to the same person. Many programs have ways to standardize references locally but the power is there to have, once and for all, one standardized author ID with all associated info. Sure, the output might still end up as "Smith, J." in some bib formats. But at least the information would be kept. And there could be author pages with a lot of info, from institutional affiliation to publication lists and professional highlights. The main advantage of having a centralised system is that changes could be applied globally (as in "across the system") as opposed to customised by each user. Authors could register themselves and add pertinent information. Readers could send comments to authors (if allowed explicitly). Copies of some publications could be linked directly. Comments by many users could linked to a given publication. Think of the opportunities for collaboration!
And the simple time-saving advantage of having, once and for all, the correct, "official" capitalization of the title.
One important point: reading notes. Bibliographies are great. The maximal information needed for a given item in a bibliography would seem quite minimal (author(s), date(s), title(s)…). Presentation/format became an important issue because some publications are quite strict in their opinion that theirs is the "correct" way to display a reference. Yet there's much more that can be done with a database of academic references.
Yes, including reading notes.
Maybe it's just a personal thing but active reading implies some type of note-taking, IMHO. Doesn't need to be very elaborate and a lot of it can be cryptic. But it's truly incredible to see how useful it can be to have a file containing all reading notes (with metadata) from one pass over a given text. With simple search technology, looking for all things you've read that made you think of a specific concept can be unbelievably efficient in bringing ideas together. Nothing really fancy. Just a list of matches for a keyword. Basic database stuff. But, oh so good!
Again, it might be personal. What I tend to is to create a file for a given text I read and write notes with associated page numbers. Sometimes, it's more about a stream of consciousness started by a quote. Sometimes, it's the equivalent of underlining, for future use. And, sometimes, it's just a reminder of what's said in the text. This type of active reading is incredibly long but the effects can be amazing.
Of course, we all use different systems. It'd just be nice to have a way to integrate these practices with reference software. And to PDAs, of course! And PDFs!
The dream: you read an article in PDF format on your PDA, you "enter" your reading notes directly in the PDF, and they're linked to your reference software. You could even share some of these notes with colleagues along with the PDF file.
Oh, sure, many people prefer to do their readings offline and few people have the inclination to type the notes they scribble in the margins. But for those of us who do most of our reading online, there could/should be ways to make life so much easier. Technologically, it should be quite easy to do.
[…Stop rambling. Well, for now, at least.]