On 1/3/12 8:58 AM, Richard Salz wrote:
[If] EXI is faster to generate, transmit, and parse than human-readable XML.... then I think we should welcome it with open arms and adopt it much more widely.I think that's necessary, but not sufficient. We also need to consider the risks of non-universal support, the burden of requiring this new system on "xml infrastructure" and "application" developers (e.g., those who work on implementations and those who would use it), the impact of Moore's law on the need for this, alternative approaches such as profiling the XML specification, and so on. Speaking just for myself, I find the comparison with XML 1.0, 1.1, and "5th edition" interesting. Liam posted that one of the reasons he supports EXI is that it reduce(s|ed) the likelihood of a variety of non-interoperable islands. Interestingly, IBM withdrew from the EXI working group because we were concerned that EXI posed a real threat of doing exactly that, with not enough consideration given to some of the items I mentioned above. This note, and reference [4] from within it, are good places to start reading if anyone is interested: http://lists.w3.org/Archives/Public/www-tag/2007Oct/0159.html I was there too. IBM's position was not well stated. And it relied on an experimental XML parser built custom for the purpose of showing how fast you could parse text XML. As far as I know, it never was released in any way and they declined to show or provide code. It seemed to rely on some tricks that would not likely be applicable to production code. My guess, purely speculation, was that IBM killed their participation based on possible product impact with their commercial development stacks or something similar. In any case, the whole episode is not a credible reason to dismiss EXI. Look for much smaller and faster XML parsers, sure. I wrote one myself last year to have a nice library for Java or Android apps: Ssx. (I suppose it is a good example of "desperate hacker" mode: Sax in a day, my own tastes in a better-than-DOM API that includes mini-XPath in a couple more. Optimized for Android/Java in 3 more days.) There are a few features that aren't in EXI that should be, IMHO, but EXI is pretty good. sdw Disclaimers: While part of the team that developed the IBM viewpoint, I don't speak for IBM. It's been a long time since I've been involved with XML core stuff at any detail. /r$ -- STSM, WebSphere Appliance Architect https://www.ibm.com/developerworks/mydeveloperworks/blogs/soma/ |