markrmiller commented on PR #2499:
URL: https://github.com/apache/solr/pull/2499#issuecomment-2159688977

   “what do you think of Jason's approach here?”
   
   It is a kind thorny practice that evades a good definitive judgment.
   
   I suppose the idea is, at some, perhaps fairly arbitrary limit, you fail 
things to force the user to reckon with what they are doing more carefully and 
actively raise limits of they think that's the correct action. I say perhaps 
fairly arbitrary because it's difficult to do any better, given varying 
hardware and data.
   
   The cost is that is that things that worked fine in testing may start 
blowing up randomly in production, perhaps needlessly, until you address the 
situation.
   Thorny. Benifit / reward, what lines are drawn where, it's all pretty 
subjective. So I don't know.
   
   As a user, I certainly would want to be able to opt in, though yes, that 
makes the effort likely of little value, particularly if it's opt-in per limit. 
A universal opt-in would be better but would likely still end up as little 
value then. As a user, I'll still be pissed when production starts failing 
things the first time, and I judge it, should not be doing that for my case. 
And at that point, I would be very unhappy if there was not a universal 
opt-out. Thorny.
   
   I do see the value, but in most cases, that value would not likely be high 
enough to convince me as a user that this was a good default system practice. 
Other than for very particular things, it doesn't seem like a common practice 
because of how thorny it is.
   
   I think some limits have a bit more clear value, though. Things that can 
build up over time and then are very difficult to address. Something along the 
lines of no limit to the number of fields on a document, where things can work 
fine for the first 6 months, but over time, you are using more and more fields 
and eventually the system just kind of seizes under the weight, and you can't 
do much but start over. Maybe hitting something much earlier that said you are 
using a strangely large number of fields, this is not a very scable approach, 
you have to change a configuration to continue down this path, would make a lot 
more sense.
   
   


-- 
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