Re: conversion rule for set-octavation/ottava with wrong version number?

[David Kastrup]

Alex Schreiber <address@hidden> writes:

> On 2017-11-01 10:51, David Kastrup wrote:
>> 
>> I don't really agree: that way lies madness.  Do we take the time when
>> something stops working reliably, or the time when it stops working
>> altogether?  I do try to keep rules "idempotent" as much as possible,
>> making it harmless to apply them multiple times in a row.  That allows
>> just trying another sweeping convert-ly run.  But this isn't really
>> possible for all rules.
>> 
>
> Thank you, David, your argument is convincing.
> But in retrospective, it would have been nice if the old compiler
> version had given a deprecation warning.

Agreed.  The usual fadeout procedure (where it's possible) of old
interfaces is

a) don't complain
b) helpful warning
c) helpful error
d) remove compatibility code

with about one major version per stage.  In practice, I don't remember
us ever doing it that meticulously if at all.

-- 
David Kastrup