Christopher Granade › Likes

Imagine future Fermats: "I have found a truly marvellous proof of this theorem, which a bug in my social software prevents me from writing down."</geeky math joke> - Michael Nielsen

I came across a similar problem a few days back although it's now fixed for me. Only way I could get links to work was to go to each item's page. - Dan Hagon

I've had similar glitches occur after my network connection drops and is re-established. - Christopher Granade

Geek in me likes 2nd comment :-)) - JoeCamel

Here's the algorithm in a nutshell: For a given hypothetical value of "e," he calculates how many charges each droplet has (a fractional number). He then calculates the total squared distance from the nearest integers to all of these values. The lower the total distance, the better the fit. This leads to a wildly fluctuating function (see above left), but it turns out that the global minimum produces the "correct" value of e! - Steve Koch

Clever. I've never seen anything quite like it. I do know of cases where people use the fact that something's an integer in a clever way; e.g., if you can bound 1.2 < x < 2.7 and you know x is an integer, then x = 2. (Come to think of it, I have a paper where we did something like that. All cleverness due to a couple of my coauthors, though!) But this is very different! - Michael Nielsen

Thanks for the feedback, Michael! I agree it's very clever. I'll let John know to look up your paper. I know he's in the midst of a topology take-home final, so I probably shouldn't bother him with too much right now :) - Steve Koch

Well I guess that's not surprising given my biases - at some level I'm more interested in what people think I've missed than my own predjudices though. FWIW I think a clever combination of DropBox, FriendFeed and some of the elements from StackOverflow, with perhaps a bit of the coordination ability of posterous would go very close to the mark. Still need better network and filter management tools though - somehow they need more configurability but less configuration... - Cameron Neylon

OpenWetWare is looking to make a major overhaul in the next couple months, and has a bit over 1 year of funding left. I feel like this is an opportunity to at least try to do some of the things that most people think are necessary for SS4S. Not perfect, but better so that we'd have a better idea of what is really needed. I think the time frame (now; already funded) makes "not perfect" a... more... - Steve Koch

I really like what you said in point 10. It's something that I've seen far too many scientists being cavalier about. Federation, open protocols and specifications, along with open source, are very important to science. - Christopher Granade

Might be worth seeing how far sourceforge meets your criteria. Certainly it's totally based around objects, i.e. software projects, and there are lots of high quality open source science projects whose code is hosted there. Although it has community/social networking tools I've personally never really used these and most visits I've had to sf have either been fleeting (to download... more... - Dan Hagon

Steve, absolutely we need to keep evolving with the resources available. OWW is a great place to do that. - Cameron Neylon

Dan, there was a conversation around using Github in a similar way some months ago and I think these things have a lot of potential as a back end. I think federation is important enough that you'd want to use a DVCS rather than SVN as a back end though. - Cameron Neylon

Sourceforge has several DVCS options in addition to svn these days. Although github is great I would be wary of anything that requires scientists to learn the intricacies of git. hg and bzr are much more friendly to non-developer types that don't need the full flexibility of git. I've had some success using them to collaboratively author LaTeX documents. - Matt Leifer

Matt, ok, I'm behind the times (nothing new there!). The intracies are less of an issue as this would only be a back end. No SS4S that any significant proportion of scientists use is going to look _anything_ like a code repository. To start with your average scientist is never going to touch a command line. If you're dealing in Latex you're already talking about a minority I'm afraid.... more... - Cameron Neylon

There are several wikis that use DVCS as a backend. This could be a starting point for developing the type of thing you are interested in. - Matt Leifer

LaTeX isn't the minority in whole areas of math, CS, physics....I guess that brings up the same old complaint: "science" is defined as all biomed, all the time. I'll try to come up with some more substantive comments though - Christina Pikas

Christina, didn't mean to say it should be excluded just that a non-command line system is non-negotiable so most online VCS aren't going to be good enough as a front end. Support for Word, Excel, video, images, XML and Latex are all non-negotiable characteristics of any such system. - Cameron Neylon

Matt, not sure that a wiki is the right starting point - the document model doesn't seem right to me, although I'm way behind on the most recent developments in Wikis so I may be out of date on that as well. What is in my head is a DVCS back end with APIs providing access from e.g document authoring systems, databases, publishers, whatever. A feed system that looks a bit like friendfeed... more... - Cameron Neylon

I wasn't suggesting actually using one of the wikis, just that they have already done a reasonable job of abstracting the version control functionality (in fact, some of them support more than on DVCS in this way) so there may be some things in the codebase that are useful. It is also an example of taking a command-line DVCS and giving it a more user friendly interface. In addition, if... more... - Matt Leifer

Ah good to know - which do you think are the best examples of these wikis? I should take a look. In any case at this stage I'm just throwing ideas out. Have no resource to actually a build anything at moment. - Cameron Neylon

Is there actually a need for social software for scientists? Or should scientists use and customize the existing social networking tools (FriendFeed, Facebook, LinkedIn, etc.)? - Martin Fenner

I'm beginning to think the main issue will be that business models for consumers services are incompatible with what researchers need. So yes, customise might be better than build but if we have to go down that route we may as well have a good idea of whats required. One person's customisation is another person's build. - Cameron Neylon

I'd be curious what you think of HubZero, Cameron. - D0r0th34

Depends a bit on server setup. For Mercurial I like Hatta, but it requires persistent python processes, i.e. no good for most shared hosts that only allow CGI. There is a list of RCS backed wikis here: http://hatta.sheep.art.pl/Similar projects - Matt Leifer

Cameron, I love and absolutely agree with the necessity of "scientific objects". If you lack those, then (as Martin points out) just use the general purpose sites. In that principle, I think there are some viable networks -- DVCS systems around scientific code, Mendeley around scientific publications, (eventually our BioGPS around genes). But I think we should be developing specific networks appealing to specific groups of researchers, rather than trying to serve the needs of all scientists... - Andrew Su

Andrew agreed, but if these are federated then they can all still talk to each other. I'm thinking more framework than site or single service. Ideally all of these things can be plugged in or wired up together...my concern with general purpose sites is primarily that they don't provide the level of trust and stability that we would expect for "research enterprise" - Cameron Neylon

Just one comment. There are protocols out there that allow different social networks to talk to each other. There are protocols out there that allow web resources to talk to each other. It's not really that hard if everyone supports some basic standards. RESTful API's, OAuth, OpeniD/Facebook Connect/Friend Connect, etc. IMO what's more important is that any sites we design have the... more... - Deepak Singh

@D only really had a chance to have a quick look. First impressions are that it is very slick but looks as though everything has to be on the inside - I don't see much mention of pulling stuff in and out. The multimedia talks are nice but why not pull them in from e.g. slideshare to pick an example. - Cameron Neylon

completely agreed, federation through standards... - Andrew Su

Twitter is far from perfect, but look at the infrastructure that has evolved around it e.g. 3rd party apps, services). You don't get that kind of traction around a social networking site just for scientists. Imagine what email or the WWW would look like if there were separate versions just for scientists. - Martin Fenner from iPhone

Absolutely but that actually means we can build something better, and as long as it hooks into Twitter (RSS/OAuth...Deepak's list basically) we get all the benefits and all of the functionality we want - as well as a way of drawing people in. Assuming this framework is any good of course. Imagine PubMed if it had been built for the consumer web (actually maybe not such a good example... more... - Cameron Neylon

Sort of responding to Deepak a few comments earlier. Something like a social network is useful for at least one reason: recruiting scientists who aren't ready for open science, or cannot communicate openly for one reason or another. So, a reasonably secure way of making data private and shared with a limited network is a good thing, I think. I think ultimately that will lead to much more open science (my own lab started out with a private wiki before doing ONS)... - Steve Koch

Steve, but does it have to be a social network per se, or a site for say sequencing geeks (I am looking at you SeqAnswers) with the appropriate features built in. Social networks don't have to be all in the open. Facebook is a social network. 90% of my communication on there is private and you should see how much of my Twitter usage is DM's - Deepak Singh