It’s just data

HTML5 Interoperability

Brian Hogan: I am getting really, really tired of Mozilla screwing up the HTML5 specification. First it’s video, and now "we’re never supporting Web SQL"

By video, I’m assuming that Brian is referring to the fact that Mozilla does not plan to implement H.264.  The story on H.264 is still a bit complicated. I’ll also add that the full story on video includes WebM, Ogg Theora, and even Flash and Silverlight, at least as fallbacks.

With respect to Web SQL, what would be needed is for somebody to define an interoperable dialect of SQL which would be support multiple interoperable implementations.  Statements like User agents must implement the SQL dialect supported by Sqlite 3.6.19 will never fly.  Additionally, not everybody agrees that stringifying SQL statements is JavaScript is desirable.


samruby: @bphogan http://intertwingly.net/blog/2010/06/27/HTML5-Interoperability (sorry for the delay, I don't read tweets much)

samruby: @bphogan [link] (sorry for the delay, I don’t read tweets much)...

Excerpt from Twitter / samruby at

Yeah, also, SQL is not a particularly great way to manage remote client-side data because instead of having to deal with just one version of your data schema, you potentially have quite a few different versions out in the wild.  That means you have to figure which version of the schema you’re dealing with and carefully migrate to the version of the schema your system actually expects.  And you potentially need to run this check every time the page is loaded.  Schema-less data stores make so much more sense for local storage in the browser.

Posted by Bob Aman at

(That or else never change your schema.)

Posted by Bob Aman at

If your data structures have changed significantly, you still need to migrate the persisted data, which is typically more tedious and error-prone in schema-less storage because the storage backend provides no tools for it.

Not to dispute that schema-less storage makes more sense for client-side storage – it does. But that’s because it embodies a different set of trade-offs than schema-based storage. It simplifies the most common scenarios (primarily the addition of columns). It does not, however, magically remove the need for migration or cure its painfulness.

Complexity in the problem space is like a waterbed, if you push it away in one place it pops up in another.

(Cf. REST constraints vs RPC everything-goes.)

Posted by Aristotle Pagaltzis at

Being a Mozilla employee, I’m biased, but for what it’s worth: Those two particular things are a part of Mozilla’s effort to ensure the HTML5 spec and the implementation of it doesn’t screw up the web.

Posted by l.m.orchard at

Add your comment