Another kind of trace: a chat with Tony Garnock Jones (of RabbitMQ fame) about intended use of XTrace

leithaus edited this page Sep 13, 2010 · 1 revision

10:09 AM me:
You might like that one
Tony: /me clicks
10:10 AM me: the main idea is to use XML schema validation as a trace validator
Tony: heh :)
so wrt XTrace
“trace” in the sense of trace-tree-of-actions-taken-by-a-process?
me: that’s correct
10:13 AM Tony: then the schema would be a kind of type for the process
me: that’s the idea
Tony: with subsequent actions being tags leading off a node
containing in turn subsequent actions
very interesting
me: that’s the idea
the aikido move here is this
Tony: … and of course there’s always relaxng :)
me: you don’t get coverage of all possible traces
Tony: (i’m wondering what kinds of xml documents are naturally viewed as traces)
10:14 AM me: but the use case is you’ve get a trace that is puzzling you
so you probe it for different properties
until you narrowed down to surprising behavior
Tony: right ok
incrementally refining the type
and checking it as you go
10:15 AM me: that’s it
Tony: til you get to a type you expect that it doesn’t inhabit
me: that’s it exactly
Tony: neat
/me . o O ( potential approaches to automatic search of the space for interesting boundary conditions )
10:16 AM me: yeah
which connects to principal types
10:17 AM hey — can i put this chat up on the project page as a kind of description of the intended use case?
Tony: sure!
me: sweet!