pytxab

Personally, the "Undo Send" feature is one of my all-time favorites in Gmail. A lot of people are comparing it to an outbox, but while the two are functionally similar, they are psychologically different. The undo send simply gives you a few extra seconds to re-read your email and catch any small errors before sending.

When the feature first launched, I was worried I would eventually adjust to it, and stop re-reading emails before clicking "send". So far, that has not happened - it's effectively a psychological trick that continues to make my emails better proof-read.

Scribd is hiring interns, both for the summer and for during the school year. We have two interns now and it's working out great.

http://www.scribd.com/jobs

JODConverter is definitely a reasonable solution for most applications. You can also write a Python script that leverages OpenOffice's Python UNO API to do the same thing - a basic script that turns Office files into PDF's is only ~20 lines of Python.

For really serious applications, though, you will eventually run in the limits of OpenOffice's support for MS Office formats. If you are very picky (or have a very picky client) about the quality fo the conversions, you'll end up having to use Microsoft Office to do the conversions. This is a lot less pleasant to set up, so I wouldn't do it unless you have to.

Before you build it, you might be interested to know that Scribd (and Docstoc, and others) have long had API's that offer this as a free service. We have seen pretty good usage at Scribd, but never enough to think that we could charge for it.

On the other hand, API's for document sharing sites are more complex and offer a lot of other functionality - this may scare off users who just want the transformation. I could see a market opportunity for a Twilio style API that was very targeted at this functionality. I still don't think you'd make a great deal of money charging for it, but I could see it getting some use.

The best way by far is wkhtmltopdf, because it uses webkit to render the page. Most other open source projects use a toy HTML renderer, which is not going to work for in-the-wild webpages.

Alexa data on Scribd is generally pretty accurate, actually. But the time on site measurement is not.

I could share screenshots of our Google Analytics time on site graph, but the hard-core conspiracy theorists would just claim we'd 'shopped them.

Doubt it. This isn't a big enough feature to be driven by that situation. It sounds like a routine product improvement to me.

So, curious what people think about this. Imagine you were running foursquare. Would you block check.in?

I could see an argument for it. If Foursquare blocked them, it would be difficult for them to gain traction. And as the front-runner in the space, Foursquare is arguably hurt more by aggregator services like Check.in, while its competitors are helped by them.

Thanks very much for the feedback! We'll investigate all of these. Opera support has become a high priority fix for us.

Whoa. That's .... not good. What OS? And you're viewing the http://www.scribd.com/html5 document?

Check it out on this baby:

http://www.scribd.com/documents/5/Paper-5

That's a LaTeX paper by one of the guys who built the reader.

Totally hear you. For the first version we focused on making it work on browsers that support @font-face. The next version will be all about making it work pretty well (if not perfectly) on browsers that don't.

That's a great point, though, thanks for letting us know how important that is to you.

Excellent point about the netbook screens. Thanks!

Actually, browsers often implement features that aren't in a standard, just because they think it's a good idea. Some of those end up being memorialized by the W3C and part of HTML. Many of the features of HTML5 started that way, including, in this case, web fonts (part of CSS3).

As the article points out, they are only using parts of HTML5 that are supported by older browsers, including IE6.

This article basically describes my experience exactly. If you have been told you have RSI, do NOT ignore this possibility.