dsmiley commented on PR #2499:
URL: https://github.com/apache/solr/pull/2499#issuecomment-2155349004

   I'm sorry for challenging your due diligence on soliciting input.
   
   > In short, my thought process was that the more broadly applicable the 
limit is, the more effective it will be at protecting stability. It's hard to 
achieve stability if the safeguards you're relying on are opt-in, "off" by 
default, or only apply to a subset of relevant queries. So I wanted to make the 
"minPrefix" limit be opt-out, apply broadly, etc.
   
   This runs the real risk of an obscure limit showing itself in production 
that would hot have been caught in a test setup for a use-case that should 
absolutely not have such a limit.  Generally "q" is provided by a user, uses a 
more user-facing query parser (namely edismax), and would be a nice place for 
such a limit.  But otherwise, queries show up in tons of other places for 
internal use-cases written by systems programmers (i.e. us devs) using the 
default "lucene" query parser where I don't think such a limit at all makes 
sense (at least not by default).  Random examples -- delete-by-query, 
facet.query, join query.
   
   The same is so for maxBooleanClauses!  I was hoping you wouldn't bring that 
up but of course you did; it's the precedent for your decision.  That 
unfortunate precedent caused a huge ugly debate with only one person holding 
veto power that insisted on maxBooleanClauses as it exists today.  That might 
have made Yonik leave to do Heliosearch over, if there was a huge straw break 
that back.  I hope we don't repeat that mistake.  I've only ever increased 
maxBooleanClauses to something insane, not because there isn't a risk of users 
doing expensive queries that we want to limit, but instead because the correct 
(IMO) place to enforce query constraints is at the query parser.  And we do 
this where I work for various things that in-effect constraint the query 
complexity.
   
   @markrmiller you and most others were involved in that ugly debate... what 
do you think of Jason's approach here?


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to