[tz] Turkey delays winter time

[Deborah Goldsmith]

If the Turkey change is still not confirmed, or is in doubt, then we should wait for further confirmation, I agree.

> And in the more usual case where a government stands pat, there is an advantage to batching changes rather than generating a new tzdata release for each individual change. Doing the latter would have doubled the number of tzdata releases this year. There is a cost to making tzdata releases, and we shouldn’t inflict that cost on everybody merely because some organizations' release processes are so slow that they'll have problems no matter what tzdata does.

First, “some organizations” correspond to the vast majority of end users using the TZ data: all the major consumer OS vendors. So to claim the current process is working except for “some organizations” doesn’t really reflect the impact on, literally, billions of people.

Part of the reason it takes those organizations a while to process an update is that, in general, when you’re pushing bits to hundreds of millions of devices you need to do thorough testing. There are also release cycles, as users don’t like to be bombarded with requests to update their devices too frequently. It’s not because we’re slow for the fun of it, or because we’re burdened with bureaucracy: the business model for consumer devices is different from Red Hat or Ubuntu.

There is little impact to client organizations from having more TZ releases. They are free to not pick up every single one. It seems to me that delivering changes ASAP is a higher priority: that gives maximum flexibility to client organizations. If there is impact on IANA from releasing more frequent updates, then we should definitely discuss it. But if it’s just counting releases, why does anyone care if there are twice as many?

Yes, sometimes governments change their minds. Then we’re pretty much hosed no matter what. And we can always back out a change before it’s released if a government changes its mind.

Deborah

> On Sep 20, 2015, at 12:14 PM, Paul Eggert <eggert at CS.UCLA.EDU> wrote:
> 
> Deborah Goldsmith wrote:
>> It seems reasonable to me to release a TZ update as soon as there’s confirmation that the change is correct
> 
> Sometimes even after we have confirmation, a government changes its mind and un-confirms.
> 
> And in the more usual case where a government stands pat, there is an advantage to batching changes rather than generating a new tzdata release for each individual change. Doing the latter would have doubled the number of tzdata releases this year. There is a cost to making tzdata releases, and we shouldn't inflict that cost on everybody merely because some organizations' release processes are so slow that they'll have problems no matter what tzdata does.
> 
> The situation in Turkey is strange. It seems that the news was reported in a major Turkish newpaper on September 8. Only yesterday were we informed. As far as we know, there's been no government announcement. The last time Turkey tried this sort of thing, it didn't take: only parts of the government observed the official time, and civil society (e.g., airlines) ignored the change. It's not clear what will happen this time, even if we get confirmation from some agency of the Turkish government. Quite possibly Apple and the tzdata project will get complaints no matter what we do. Let the fingerpointing begin!