It's interesting to see how XML was shoehorned into everything regardless of whether it makes sense and then sold as "jump on the bandwagon or get left behind", and see the similarities to later hype-cycles such as when everything had to be blockchain.
Yes, you were supposed to publish your invoices or whatever in a specific schema, then send them to another company who would use XSLT to transform it into their schema, and because everyone was very smart and knew how to design and publish schemas and transforms, everything would work.
In practice, BigCo published whatever the hell non-conformant XML slop generated from their internal CSV source and if you were on the receiving side you had to do all sorts of black magic to ensure your XML parser didn't die in the middle the ETL toolchain because BigCo never bothered to escape ampersands or whatever. If there was a schema it was years out of date and no-one bothered to validate against it, presumably because the validation step failed because the only one who knew how to fix it had long since left the company.
the author proudly repeats while posting falsehoods ...
Then has this line/quote:
YAML™ (rhymes with “camel”) is a human-friendly, ...
Pot, Kettle, fucking black?
The objectively true and undeniable expansion of the YAML acronym is Yet-Another-Migraine-Looming.
It's human friendly in the sense that razor wire is human friendly: you won't be harmed if you don't fucking touch it, and will absolutely work as an effective way to keep other people away from you.
Unquoted strings don't seem that valuable to me – except for map keys, which typically act as identifiers (and often map to identifiers in programming languages, which have restrictions placed upon their form) and therefore tend to be short, free of whitespace, and limited in character set.
It would be interesting to see if there was discussion about not allowing unquoted strings, or on map keys being restricted to "identifiers" instead of arbitrary values.
max-headroom | 15 hours ago
It was, in fact, the YAML spec's fault.
zaphar | 15 hours ago
It is 100% the yaml specs fault.
minus | 11 hours ago
If in doubt, blame YAML spec
fazalmajid | 9 hours ago
That take is correct, but it's like shooting a fish in a barrel. Or a xenomorph in a barrel.
sjamaan | 2 hours ago
Ouch, that would be messy!
veqq | 7 hours ago
Wow, I had no idea
mira | 59 minutes ago
Here's a good example: https://public.dhe.ibm.com/software/integration/datapower/library/DataPower_datasheet.pdf
It's interesting to see how XML was shoehorned into everything regardless of whether it makes sense and then sold as "jump on the bandwagon or get left behind", and see the similarities to later hype-cycles such as when everything had to be blockchain.
gerikson | 48 minutes ago
Yes, you were supposed to publish your invoices or whatever in a specific schema, then send them to another company who would use XSLT to transform it into their schema, and because everyone was very smart and knew how to design and publish schemas and transforms, everything would work.
In practice, BigCo published whatever the hell non-conformant XML slop generated from their internal CSV source and if you were on the receiving side you had to do all sorts of black magic to ensure your XML parser didn't die in the middle the ETL toolchain because BigCo never bothered to escape ampersands or whatever. If there was a schema it was years out of date and no-one bothered to validate against it, presumably because the validation step failed because the only one who knew how to fix it had long since left the company.
icefox | 7 hours ago
This sounds like a case of "parse, don't validate" iiuc?.
stephenr | an hour ago
The blog post starts out with (emphasis mine)
Then has this line/quote:
Pot, Kettle, fucking black?
The objectively true and undeniable expansion of the YAML acronym is Yet-Another-Migraine-Looming.
It's human friendly in the sense that razor wire is human friendly: you won't be harmed if you don't fucking touch it, and will absolutely work as an effective way to keep other people away from you.
mira | 53 minutes ago
Good job actually scouring the emails.
Unquoted strings don't seem that valuable to me – except for map keys, which typically act as identifiers (and often map to identifiers in programming languages, which have restrictions placed upon their form) and therefore tend to be short, free of whitespace, and limited in character set.
It would be interesting to see if there was discussion about not allowing unquoted strings, or on map keys being restricted to "identifiers" instead of arbitrary values.