Github user phrocker commented on the pull request:
https://github.com/apache/accumulo/pull/103#issuecomment-222277916
Josh, aka Josh, sent me a note to take a look. Because Josh, aka Josh,
tended to my chickens while I was away, will take a moment away from tending to
my flock to comment.
The paradigm you've chosen is a separate API. I understand the intent, but
I believe this not only increases the footprint of the API, but long-term
increases the costs for its maintainers ( you! ). Those who wish to implement
the layer of access you've allowed will do so without regard to an external
API. What we needed and implemented was a way to bypass the tserver. I would
like to see this supported through the connector as an established mechanism
for data retrieval.
Having a scanner to do that that, accepting the current API was most
advantageous for us, long term. It decreased the maintenance footprint, kept
the API clean and helps to abstract the tablet location obtainer algorithm.
Invariably, on a production system, we will need to cache locations of rfiles,
and re-build as compactions occur since we don't inject the scan column. We
want to hide all of this. I would love to see all of that implemented for us so
we don't need to continue to implement what we currently have. I was hoping
your pull requests would allow us to that. Perhaps i'm reaching for what I
hoped the intent of your pull request was.
Back to my chickens, and apparently now injured geese.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
