Modern XMPP project discussion - 2026-05-02

Personally, I prefer simple binary formats.

0xXX Size of Datum ID 0xXX[X...] Datum ID 0xXX Size of length number 0xXX[X...] Length of data data ...

Separation of metadata and data is a solved problem. Declare a length in bytes, it's that simple. If you want, you can even use a number to refer to the data to decouple the formatting completely from the 'content'.

It should be noted that metadata is content too, so this boils down to two bad practices neither XML or Markdown solve: - Data being referred to has no length or location constraints and therefore no clear boundaries without being parsed from beginning to end - Because metadata is special compared to content, rather than also being a form of content, the interface between the metadata and the consumer is ass cancer because the user isn't "supposed" to interact with it, except when they are. Confusion between what is metadata and what is content causes unusual errata.

This could be solved by simply keeping track of what things are. It feels almost obscene to say out loud but it's the truth.

Any issue from unclear parsing can be solved by simply not using parsing to determine what something is, and use a length field and a type number instead. You say exactly what it is and where it is and that's it.

But instead we do all this crazy parsing, and then the parsing fails to match what we meant because it has room to misinterpret the boundaries and therefore contents of data.

Because we couldn't bring ourselves to just tell it clearly where the data is.

Without any parsing.

By using... a number.

Wild stuff.