@abyrd says the following:

http://www.normes-donnees-tc.org/wp-content/uploads/2018/10/Proposition-Profil-SIRI-Lite-initial-v1-3.pdf
See section 2.1.3. This profile was a back-formation from practices at the NY MTA and OpenPlans described at https://bustime.mta.info/wiki/Developers/SIRIIntro

At today's meeting @leonardehrenfried mentioned that no one seemed to know for certain whether SIRI Lite was supposed to return JSON or XML.

My understanding is that SIRI Lite was proposed as an official aspect of SIRI by Christophe Duquesne in this document: http://www.normes-donnees-tc.org/wp-content/uploads/2018/10/Proposition-Profil-SIRI-Lite-initial-v1-3.pdf
It states that SIRI Lite returns either JSON or XML depending on the situation. See section 2.1.3 for a description of how XML is mapped to JSON.

My understanding is that SIRI Lite was a formalization of practices seen in the wild at the NY MTA and OpenPlans. This page describes the system that introduced the practices that were formalized as SIRI Lite: https://bustime.mta.info/wiki/Developers/SIRIIntro

It says: "While the SIRI standard specifies XML as an interchange format, it incurs a substantial penalty in data-transfer and processing requirements. In contrast, JSON (Javascript Object Notation) is a lightweight data-interchange format. Human-readable, lightweight, and easy to parse, JSON offers an alternative. Thus, MTA Bus Time provides both XML and JSON versions of its API."

This page asserts that the MTA contributed directly to the normalization (standardization) process:
http://www.normes-donnees-tc.org/format-dechange/donnees-temps-reel/

I updated the title and a comment to say "SIRI Lite" instead of "Light". All sources I could find spell it "Lite". The motivation is that I was baffled when searching for this ticket using the word "Lite" and couldn't find it :)

I reject this Americanism :)

I don't even know @leonardehrenfried where you found the term "Light" 😂

@leonardehrenfried I fielded a question today about retrieving SIRI-Lite from a JSON-only endpoint. Currently it looks like the updater, even the one I modified to use the HTTP GET instead of POST method, can map only XML responses. It seems like JSON support might be a common need.

So far I managed to talk people out of it.

I see that there are some rules about converting XML to JSON in the draft but I much prefer having a proper XML schema to validate the data to a French-only document that doesn't contain tooling for automatic validation.

I don't even know @leonardehrenfried where you found the term "Light" 😂

Just in case you genuinely don't know: "Lite" is an intentional commercial/marketing misspelling of "light". It's definitely not the norm to use the "lite" spelling in a serious context like a data specification. Doing so is a bit tongue-in-cheek, like this is a "new and improved" product being promoted with a kitschy late-20th-century vibe. So someone who heard it in conversation would probably assume "light" while someone reading the spec might remember the distinctive "lite" spelling.

I reject this Americanism :)

Clearly this spelling is not American, for example used on this French sauce salade! /s