RegTAP-1.1 & ADQL-2.1 - the ILIKE concerns

[Theresa Dower]

As the implementor of one of the two RegTAP 1.1 reference implementations, I support this reversion to requiring ivo_nocasematch rather than lock in the ADQL version dependency.


It is worth noting that because of the configuration of the STScI full searchable Registry database, all text matching is done case insensitively, and our implementation on this point is dodgy: both ILIKE and the user function are aliases over the default "LIKE" behavior. (So really, we implement ILIKE and ivo_nocasematch correctly, but never did for LIKE.) I should have perhaps taken this as a sign to pay more attention to this part of the spec rather than less. Thank you Marco for bringing it up.


As Reg chair, I know the voting dates for this remain undefined; we were hoping for a quick tally with the comments and bringing it up for TCG/Exec approval in November. Once the document reflects a general agreement on this issue, I'll formally set those dates and ask again for input. It seems very likely we can still make that deadline.


--Theresa


________________________________
From: registry-bounces at ivoa.net <registry-bounces at ivoa.net> on behalf of Mark Taylor <M.B.Taylor at bristol.ac.uk>
Sent: Thursday, August 8, 2019 5:02:12 AM
To: Molinaro, Marco
Cc: IVOA Registry; IVOA DAL
Subject: Re: RegTAP-1.1 & ADQL-2.1 - the ILIKE concerns

Marco,

I am all in favour of your suggestion.  As far as I can see the move
to ILIKE in RegTAP is really just a bit of tidying up that can harmlessly
be omitted or deferred if the interaction with ADQL 2.1 is potentially
problematic (which it seems it may be).  I'm not aware that ILIKE does
anything that ivo_nocasematch does not.  From an implementation point
of view in topcat I'm unlikely to be using ILIKE in this context
(except maybe for example queries copied from the RegTAP 1.1 document),
since ivo_nocasematch is required for all RegTAP service versions
and ILIKE is not, so there's less logic required to just use the
UDF in all cases.

Mark


On Thu, 8 Aug 2019, Molinaro, Marco wrote:

> Dear ReR-ers,
> (DAL-ers for reference)
>
> in the process of reviewing RegTAP-1.1 PR from the DAL perspective
> a consideration come up with respect to the strong dependency of
> RegTAP-1.1 case insensitive comparison on ADQL-2.1 (PR itself)
> ILIKE operator.
>
> Two issues came up: RegTAP waiting ADQL for proper dependency
> reference; too strong dependency/forcing of RegTAP requirement
> over the ADQL specification.
>
> The first is actually minor/can-be-worked-around, the latter is more
> critical.
>
> As things currently stand, there's the risk RegTAP controls ADQL behaviour
> of ILIKE or that ADQL can break RegTAP when/if changing ILIKE definition.
>
> Proposed solution is to re-instate the ivo_nocasematch in RegTAP,
> decoupling the specifications and releasing the two issues above in one
> shot.
>
> Comments and discussion highly welcome (including detailed explanations
> from the spec editors and thoughts from implementors).
>
> Cheers
>       Marco
>

--
Mark Taylor   Astronomical Programmer   Physics, Bristol University, UK
m.b.taylor at bris.ac.uk +44-117-9288776  http://www.star.bris.ac.uk/~mbt/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ivoa.net/pipermail/registry/attachments/20190809/eeb4178b/attachment.html>