I don't think "machine readable" needs to be defined any more clearly for now—it's a SHOULD, so if you wanted to ignore the rule you'd be perfectly fine anyway.
What are the rules for determining the source of a message? Could we have a web-based form for transferring currency that simultaneously updated the records and sent an email to the PF "in the name of" whoever was doing the transferring? Gaelan > On May 16, 2017, at 9:46 AM, Kerim Aydin <ke...@u.washington.edu> wrote: > > > >> On Tue, 16 May 2017, Alex Smith wrote: >>> On Tue, 2017-05-16 at 09:14 -0700, Kerim Aydin wrote: >>> That said, if an officer figures out a way for some transactions to be truly >>> automated (e.g. enter a transaction on a website and it gets logged and sent >>> to the PF), I'm all for that. >> >> I've long thought that if an action has a sufficiently well-defined >> format, we should define a machine-readable format for it and have that >> action taken by posting that machine-readable string to a PF, rather >> than by announcement; this is mostly to make it unambiguous whether or >> not there's an attempt to take the action. That way, you could use a >> computer program to handle office reports and the like. > > Not that I want to add an Office, but this could be done through a maintained > specification: "The Automator maintains a document that contains a list of > synonyms for actions; thse can be changed by (whatever method)". Any > part of a public message enclosed in &%&%&% is taken to apply that > translation. > > Example entry on a synonym list: > A statement reading "TRANSFER,Currency,name1,name2,N" > is a synonym for "I transfer N units of Currency from name1 to name2". > > An issue I see is things like fluidity of translation; it would require that > we > be VERY strict on things like nicknames. If someone entered something like > that > but left the period off my nickname (G instead of G.), it would be clear and > transparent to any human but would break a machine. I don't think we'd want > to be that strict, so the expectation would be that a machine could scan and > find these transactions, but the human would always have to monitor and > confirm/ > edit each entry. > > > > > >
