<div class="gmail_quote">On Fri, Jun 19, 2009 at 11:53 AM, Steve Hosgood <span dir="ltr"><<a href="mailto:steve@tallyho.bc.nu">steve@tallyho.bc.nu</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">80n wrote:<br>
> There's a problem with the minute diffs, since 0.6, caused by large<br>
> changesets. The workaround is to delay them by 30 minutes, but it's<br>
> possible there was missing data before the problem was fully diagnosed.<br>
><br>
> It's a bit of a moot problem at the moment as the main XAPI server is<br>
> dead. Any new servers that come up (some are being worked on) will<br>
> start with a clean planet so things should be much better.<br>
><br>
</div><div class="im">Ah - so you're saying that XAPI servers work from some historical planet<br>
file which has subsequently been modified by applying zillions of minute<br>
diffs from that point in history onwards are you?<br>
<br>
So if the diff carrying info about those missing roads had got lost<br>
months back, those ways will remain missing on XAPI for ever (or until<br>
someone modifies those ways again which would presumably make them<br>
appear in a fresh minute diff).<br>
<br>
If this is so I could imagine it would be a good idea for each XAPI<br>
server to (randomly) load itself a fresh, clean planet file from time to<br>
time (say once a week or once a fortnight).<br>
</div></blockquote><div><br>Steve<br>Yes, a full resynchronisation does get performed periodically, but not yet routinely.<br><br>There was some discussion a while back about maintaining record counts or checksums or something so that the integrity of any mirror database (XAPI, TRAPI, ROMA, etc) could be verified. It's a good idea but I don't think anything has been done yet.<br>
<br>80n<br><br><br></div></div><br>