On 5/31/20 12:09 PM, Florian Weimer wrote:
* Paul Eggert:


On 5/30/20 7:00 AM, Florian Weimer wrote:

Would this be a good opportunity to engage with the government of Morocco
to encourage them to give more warning?

My understanding is that at present, they don't know themselves in
advance.

Yes, I think you're right.


They would have to redefine the criteria to something based
exclusively on astronomical calculations

That shouldn't be necessary, as they're using a conservative
approximation; that is, it's OK if their approximation is a bit off
because all they need to do is to bracket the actual Ramadan rather
than predict its Gregorian dates exactly.  If they had an algorithm
they could publish it and we could use it. And it would be
technically feasible for them to have an algorithm; see the
Morocco-prediction code in the tzdb 'africa' file for an example
algorithm.

It looks like the algorithm predicted correctly the data of Eid
al-Fitr as 24 May, but the derivation of the end date from that is
suspicious.  I think it would make more sense if the time change does
not happen on Eid itself.


In Morocco (where I live), the end of Ramadan (arabic month) is follow by the Eid al-Fitr, and concretely it's 1 or 2 day off for the people (with traditional visiting of family, big lunches/diners, etc.). So for this year the astronomical calculations don't include the following 2 day off in the calc. This 2 days have fall in a Sunday/Monday, so it's not acceptable by people to have a time shift during these 2 day off.

Perhaps, you can modify the (predict) rules for next years : if the end of Ramadan is a (probable) Friday or Saturday (and so the 2 day off is on a weekend), the next time shift will be the next weekend.




 A rule like this would ensure that Eid
still uses the same time as Ramadan in every year, which is likely
what people expect.