- From: Robin Berjon <robin@berjon.com>
- Date: Mon, 29 Aug 2011 12:30:03 +0200
- To: Brian LeRoux <brian.leroux@nitobi.com>
- Cc: Marcos Caceres <marcosscaceres@gmail.com>, public-native-web-apps@w3.org
On Aug 26, 2011, at 22:20 , Brian LeRoux wrote: > brutal! we raised all those issues w/ the DAP. I'm unsure if they > listened but the point is a spec is useless w/ a plausible > implementation and if the implementation requires a quirks section > that is a page long then perhaps the spec needs work!!! I don't think that there's any doubt in anyone's mind that a spec is useless without implementations (and won't be good without implementation feedback during its development). The specific problem with the Contacts API is that it has to somehow expose what is by nature very messy data, implemented messily. A lot of the quirks you list are for information items that the underlying platforms don't provide. This opens a number of interesting questions: • Is it better to expose a field that is present in the majority of platforms but that will always return null/false in some cases, or should it simply never be available until all platforms support it fully? My personal take here is "it depends (on the usefulness of the field, on the disparity of data, etc.)". • Should the specification require that implementations make some form of effort to normalise the mess that are underlying contacts databases at the risk of increased complexity and of introducing weird behaviour we'll want to get rid of when contact DBs get better, or is it more useful to expose them as-is and leave it up to authors to use libraries that can clean things up? I think it's more useful to expose the chaos, a spec isn't a great place to plaster over such things. • Does it make sense to create a read-only API for v1 considering that all the difficult problems are in successfully round-tripping information? That's very much an open one I'd think. Those are certainly things I'd love to see discussed! Either way, I don't think that the above list's "brutal" about the Contacts API as much as it shows the brutality that the Contacts API has had to deal with in trying to be usefully specified :) -- Robin Berjon - http://berjon.com/ - @robinberjon
Received on Monday, 29 August 2011 10:30:39 UTC