On 24 Apr 2012, at 11:10, Stéphane Ducasse wrote: >>> And yet again I point to Tirade :) >>> >>> http://goran.krampe.se/blog/Squeak/Tirade.rdoc >>> http://goran.krampe.se/blog/Squeak/Tirade2.rdoc >>> http://goran.krampe.se/blog/Squeak/Tirade3.rdoc >>> >>> Especially Tirade2 above shows a bit about size (4 classes, 500 loc) speed >>> and portability. Tirade is basically a parser for Smalltalk messages that >>> only are allowed to use literals as arguments (although arbitrarily nested >>> literals). >>> >>> Which is exactly what Stef describes + a bit more. :) >> >> Yeah, I remember reading that a long time ago. It is indeed a cool idea, >> Göran. Reminds me of the Erlang related UBF >> (http://www.sics.se/~joe/ubf/site/home.html). >> >> I think the JSON choise is not bad: it is simple and universally accepted. > > But you can express **EXACTLY** the same with > #( > > > ) > > So what is the point? > > Stef
Yes, you are right, they are technically mostly equivalent. (JSON has simpler primitive types, clear escapes, lists/arrays and maps/dictionaries). But the point is, there are so many formats out there, and everybody likes to make there own. If you pick JSON, the discussion ends. It is an RFC standard. If you pick something that looks suspiciously like some (for most people) weird programming language you will get discussions, always. Dale said so: it is a pragmatic choice. Now, given the fact that the domain here is Smalltalk anyway, there is something to say for using a Smalltalk based representation. But then you need to write a clear spec and a non-compiler based parser. With the JSON meta data, you could envision other non-Smalltalk tools using it more easily. Sven -- Sven Van Caekenberghe http://stfx.eu Smalltalk is the Red Pill
