Hacker News new | ask | show | jobs
by unalone 6302 days ago
I'm going to be a devil's advocate.

1. It's successful, at least in terms of traffic, so it's worth watching and perhaps learning from.

The lesson here is that if your concept is extremely general, you'll have success. Scribd went after a niche that was completely underused, and props to them: they got just as big as that plan would predict. That said, it's still not a good web site. It's not a particularly good role model for people who want to build their own web site, unless said people have an idea that's just as large, and large ideas are hard to come by.

2. The web hasn't obliterated the old document formats (yet). Not everyone that has content in those formats has the time or inclination to replicate the layout in HTML, so Scribd is making distribution easier for them. It may not be solving your problem but it is solving some people's problem.

It's solving the problem, but again: not well. Google's suite of PDF-to-HTML and slide displayers do a much better job of this than Scribd's iPaper does. Why that system hasn't been used yet for a public site I don't know: perhaps Google's scared of getting supersued again.

As I said on the other thread (http://news.ycombinator.com/item?id=521565) the real challenge is making PDF displays convert to HTML. I'm certain it's possible: hell, with SIFR you could even keep the PDF's embedded fonts. That's a really interesting challenge (and it's one that I'm working on, though its implementation will be radically different than Scribd's much more all-purpose viewer). The end result is that you'll get something that requires no custom displays, because it just works the first time.

People here usually have a soft spot for YC funded startups; not sure why Scribd's treated like the disowned member of the family.

We encourage YC startups. Note how rarely we attack DocStoc, which is way, way worse than Scribd is. However, HN isn't a site of happy encouraging people. We're all critics of everything, which is awesome. There are also pretty vehement Reddit critics, a few J.tv critics, a lot of people who make fun of OMGPOP, and on and on. So think of this like tough love. We'd love to love Scribd, but we won't tell it we do until it rehabs itself up. (I also suspect there's some jealousy of its success, along with some annoyance that they don't need to take our advice, good as it is, because they're a behemoth without our help.)

1 comments

I hope you're aware of sIFR speed issues when you're replacing multiple different instances of text. It dramatically slows down the browser. (And when dealing with PDFs, I wouldn't be surprised if on average you had to use 20+ sIFR flash objects.)
What do you mean, different instances of text? Do you mean different types of text that get rendered differently? Or does this apply also when you work with a large block of text that's all rendered a single way?
Basically, each block tag (h1, h2, h3, p, li, etc.) would create a new instance of sIFR, thus creating a new flash object on the page. For example, you can't use just one flash object for a long stream of text, if that stream of text is divided into paragraphs, because sIFR cannot internally give you the necessary margins or tabbing required for marking paragraphs.

The only case where you would have one flash object is when you have a single column of text which has the same font, same leading, and no margins throughout. Unfortunately, that won't happen very often.