obs_id is related to an observation: it is identical for all granules related to the same observation, containing different type of data (e.g.: raw and calibrated data, associated geometry, etc). In many EPN-TAP v1 services, such products were described together on the same line with a unique index parameter.
All such IDs can be arbitrary alphanumeric strings.
As a general rule, different products related to the same observation are no longer described together on a single line of the epn_core view, but on successive lines associated by the same obs_id, each with a different granule_gid (and a specific granule_uid, see Table 2 below).
Each line in the epn_core view must describe only one product (plus a thumbnail wherever relevant). The notion of "main product" (more or less explicit in v1) is therefore deprecated, and the epn_core view in v2 includes more lines than in v1. Although less compact that the previous table presentation, this list presentation is much more efficient for machine-handling.From our initial discussions, all such IDs can be alphanumeric strings, TBC (in particular for granule_uid, but this seems OK with ADQL)
3) The notion of table/service parameters is deprecated