clarify KPI link checking

[tomkralidis]

Given WCMP2 KPIs will be part of the WIS2 Guide (target is INFCOM-4), coupled with clarifications of how GDCs shall emit metrics on a daily basis, per the below PR (target is FT2026-1):

wmo-im/wis2-guide#211

...this means that GDCs will (likely) run KPI checks on a daily basis and return metrics accordingly.

Some KPI checks involve testing links in a WCMP2 record, specifically:

• 2.4. Graphic overview for metadata records
• 2.5. Links health

..which means daily checks on links by GDCs. Note that all other KPIs proposed do not have network/connectivity requirements.

Is this a problem of elevated traffic, especially if a WCMP2 record has numerous HTTP links? Or potentially long running jobs as GDC content grows over time? Options:

1. keep as is; daily link checks contribute to better quality metadata with less "link rot"
2. make link checking a recommended check that a GDC performs (would need update in the Guide)
3. other options?

Thoughts? cc @wmo-im/tt-wismd

[josusky]

My order of magnitude calculation for the worst case scenario is:
1000 records * 5 URLs each * 30 sec timeout = 150000 seconds (almost 42 hours)
Of course, the best case is in order of minutes. What we can do, in pywcmp is to use asynchronous URL checking, opening tens (hundreds?) of connections at the same time, thus reducing the total wall clock time of the check by an order (two?) of magnitude.

[tomkralidis]

+1/makes sense to implement in pywcmp (associated PR in World-Meteorological-Organization/pywcmp#155). Having said this, should we be concerned by potentially "annoying" various servers/services? We know that while the WCMP2 record is on Global Service infrastructure, links will not be. The reality is then that all links in a WCMP2 record would be health checked on a daily basis for the long term. Of course, given a WCMP2 record is always public, publishing of any links in same comes with a certain realization from the data producer.

In summary, sounds like this is not a major concern unless others have concerns.

[tomkralidis]

TT-WISMD 2026-04-15

• we should consider deduplication of links (implement set math in pywcmp [@tomkralidis, @ferrighi to review])