the tragedy is we are gonna have a generation of "engineers" (if we can call them engineers) who won't know how to properly use tools such as Cassandra, ScyllaDB, FoundationDB etc.
excellent db's meant for very specific use cases. abused by resume driven mercenaries - but used with precision by missionaries.
Most devs don't understand how to use a traditional SQL style RDBMS appropriately either. Especially with support for JSON encoded fields, you can do a lot of things that break traditional normalization but have a massive impact on performance and scalability. The massive increase in hardware alone has shifted the balance a lot.
There's still lots of room for specialization tuned database options like Cassandra, Scylla, etc. I do wish more developers had a lot more awareness of how/what/where these things are appropriate.
FWIW, you see similar efforts to shove GraphQL into application stacks and environments where you gain very little benefit. Often and especially when you have the same people developing the app and managing DevOps with smaller teams.
It would certainly be a boon if Cassandra did better ACID than everyone's favourite SQL DBs (which are unknowingly running on read_committed).
It's a good sign that they've got bank account balance updates in the example code in the Accord github (sibling comment), as that's what I reach for if I suspect my DB is only pretending to be ACID.
I find it unsettling that every developer is asked for their opinion on LLMs and vibing, because of the dynamic it sets up with the pace of change in AI systems: if they give specifics about things that their experience shows LLMs can and cannot do, those are out of date immediately after publication, if not already at interview time, but if they are vague, it sounds like they have no experience with it, but also if they note the curve and try to adjust for those things, they come off like marketing for frontier orgs and it's still going to be laughable in six months or a year.
If you want to make the interview about that, fine -- there's lots of material and interest. If the interview isn't primarily about that, don't lay this rhetorical trap for the interviewee.
When I find speakers for NYC Systems I do ask them not to speak about their experience with AI (at least that the talk should not be entirely about this) because of exactly this risk of being distracting and immediately out-of-date.
However, I think it can be a useful signal to understand if/how experienced developers in major software infrastructure projects are using AI today. So I plan to keep asking the question.
Branimir is an engineer's engineer. Excellent choice.
His work is very cool and I was impressed by the thoroughness and thoughtfulness of his responses.
the tragedy is we are gonna have a generation of "engineers" (if we can call them engineers) who won't know how to properly use tools such as Cassandra, ScyllaDB, FoundationDB etc.
excellent db's meant for very specific use cases. abused by resume driven mercenaries - but used with precision by missionaries.
Most devs don't understand how to use a traditional SQL style RDBMS appropriately either. Especially with support for JSON encoded fields, you can do a lot of things that break traditional normalization but have a massive impact on performance and scalability. The massive increase in hardware alone has shifted the balance a lot.
There's still lots of room for specialization tuned database options like Cassandra, Scylla, etc. I do wish more developers had a lot more awareness of how/what/where these things are appropriate.
FWIW, you see similar efforts to shove GraphQL into application stacks and environments where you gain very little benefit. Often and especially when you have the same people developing the app and managing DevOps with smaller teams.
Not if I can help it!
Adding ACID to Cassandra sounds very interesting, will be curious to read more about that.
Probably the best single resource is https://github.com/pmcfadin/awesome-accord
Cassandra needs ACID like a fish needs a bicycle ,-)
"When is "ACID" ACID? Rarely." - https://www.bailis.org/blog/when-is-acid-acid-rarely/
It would certainly be a boon if Cassandra did better ACID than everyone's favourite SQL DBs (which are unknowingly running on read_committed).
It's a good sign that they've got bank account balance updates in the example code in the Accord github (sibling comment), as that's what I reach for if I suspect my DB is only pretending to be ACID.
The Accord developers mostly work at Apple so it would seem Apple at least wants/needs ACID transactions for Cassandra.
I find it unsettling that every developer is asked for their opinion on LLMs and vibing, because of the dynamic it sets up with the pace of change in AI systems: if they give specifics about things that their experience shows LLMs can and cannot do, those are out of date immediately after publication, if not already at interview time, but if they are vague, it sounds like they have no experience with it, but also if they note the curve and try to adjust for those things, they come off like marketing for frontier orgs and it's still going to be laughable in six months or a year.
If you want to make the interview about that, fine -- there's lots of material and interest. If the interview isn't primarily about that, don't lay this rhetorical trap for the interviewee.
When I find speakers for NYC Systems I do ask them not to speak about their experience with AI (at least that the talk should not be entirely about this) because of exactly this risk of being distracting and immediately out-of-date.
However, I think it can be a useful signal to understand if/how experienced developers in major software infrastructure projects are using AI today. So I plan to keep asking the question.