Hacker News Reader

Comment by monocasa | Hacker News Reader

Probably because Oracle has been pissing off their bread and butter for years and it's starting to bite them.

Amazon got the gov cloud contract. Oracle's suing, but all they have is that someone jumped from the DoD to Amazon. https://www.nextgov.com/it-modernization/2019/02/pentagon-in...

Also, I've heard stories at other .gov datacenters where they're trying to phase out Oracle as fast as they can because they're tired of their crap. Like last time I checked Oracle DB charged based on the number of CPUs you _could_ be running on. So if you have a modern env that dynamically provisions across a datacenter, then they charge for all of the cores in your datacenter.

View on HN

We got hit by that.

Oracle used to charge us per physical core on our production nodes, with a discount for development and backup/spare.

Then suddenly they changed how it was licensed and our bill jumped up 4x (we're talking a six figure increase).

We negotiated it down somewhat (closer to a 2.5x increase as a "goodwill discount"), but since then have been quietly making changes to the code-base to make a future SQL Server migration easier. We have four years left in this contract, and we hope to be ready by then.

We're on Oracle at all because we used to use their application development suite (ADF, Forms, etc), but those all either got discontinued or turn so sour you don't want to be using them, so now Oracle DB is just a remnant.

If you don't mind me asking, how is the SQL Server as far as price goes? I know Oracle is very expensive, is SQL Server cheaper for a similar usage pattern?

Also curious, what made you decide not to use PostgreSQL (or EnterpriseDB if you need entreprise solution)? Was it a technical reason or are you more comfortable with MS solutions in general?

I am asking because I have mostly used MySQL (pre-Oracle) and now PostgreSQL, and have always wondered what made people choose either Oracle or SQL Server. I have met both in different jobs - I hate Oracle with passion because of their licensins shenanigans, and wasn't impressed by SQL Server (though that might be in part that I was using it with SharePoint, which was... something).

SQL Server now uses the same licensing strategy as Oracle in virtual environments.

Citation needed

SQL Server offers two different license models (at Enterprise tier). Per physical core or per virtual core.

If you license all the physical cores you can run as many copies of SQL Server as your infrastructure will allow.

If you want to run SQL Server as a partial part of a private cloud then licensing per virtual core could be a significant savings in some circumstances. This is because physical cores used for other things won't require a license.

This article has more info:

https://blogs.flexera.com/elo/2017/06/an-introduction-to-mic...

> If you don't mind me asking, how is the SQL Server as far as price goes?

You can use this to get a very ballpark idea:

https://wintelguy.com/mssql-ent-licensing-calc.pl

From actual quotes at full price, we'd be paying around the same to Microsoft for SQL Server Enterprise than we were paying Oracle BEFORE the price hike/license changes. However that doesn't include discounts which most organizations running any Microsoft software at all would be eligible for.

So in other words: Almost always cheaper than Oracle Database. Varies between a little cheaper and substantially cheaper. Both companies are trying to drag you towards Cloud however, so there's deeper discounts there right now (it's a trap!).

> Also curious, what made you decide not to use PostgreSQL (or EnterpriseDB if you need entreprise solution)?

No one person decided. This is a large organization that moves slowly and is fairly conservative. Plus the costs of migration are high, if migration fails people in management will suffer (either to their career or just outright be transferred/fired).

So in other words "nobody gets fired for buying Microsoft" or w/e. It certainly doesn't hurt that it is much easier to hire/contract in SQL Server talent than PostgreSQL. Technical considerations play a part, but only up to the point where it has to have what we need.

> I am asking because I have mostly used MySQL (pre-Oracle) and now PostgreSQL, and have always wondered what made people choose either Oracle or SQL Server.

That's because you're considering what is best using logic, reason, and technical facts. In the enterprise space more often than not major companies just follow the crowd, with a handful of trailblazers setting the standard. Oracle and SQL Server are popular because they're already popular, in a fully circular way.

If I started a business tomorrow I'd use PostgreSQL. But if I was a C-level in an enterprise and my career is on the line, I am going to be sticking with SQL Server.

They also get picked because they have excellent vendor support behind them. As well as third-party ecosystems.

This. If you are c-level you aren't touching the code, but you like the idea of paying the people who built the thing a support contract so you can call and yell at them when you misuse the product and it falls over.

The alignment between product support and product builder is muddier for c-levels to follow.

> Also curious, what made you decide not to use PostgreSQL (or EnterpriseDB if you need entreprise solution)?

Keep in mind that MSSQL enterprise comes with a lot of tools beyond just a RDBMS. That was always the hangup when I have tried to move companies from MSSQL to something like PG.

Can you give example of several useful tools in the MSSQL garden?

Sure. Two of the main ones I used were MSSQL Analysis Services and Integration Services.

I built and managed a data warehouse/internal reporting system for a fairly large telecom. I used ~100 ETL packages built in SSIS to pull in data from systems all over the company, SSAS to process the data and SSRS to build reports. It worked pretty well at the time 10+ years ago.

I sense Oracle soil will collapse quite fast. JVM speed can only keep people interested so far .. and a lot of smart people are into language and compilation these days.

From my observation it’s much easier to recruit Java devs than Node devs, for example (I’m located in an average city, not a leading one like SF or NY). I’ve closed my Node projects because it seems to be too early to recruit and train interns in this area.

Being able to easily recruit programmers in a popular stack says nothing of their quality as professionals though.

If anything, I've seen a lot of non-programmers disguised as Java programmers. Armies of them in enterprise teams. A truckload of expense for not much gain.

I'm not saying it's everywhere like that of course. I've met brilliant Java developers when I contracted for SAP.

So how do you deal with such a risk?

If I had to hire I'd search programmers of less popular languages -- and only 3-5 of them. These people tend to be insanely productive and have very good perspective and culture.

It’s a running joke in companies that Hibernate/JPA helps us abstract from the database « in case we change DBs », and how it never happens because we’re tied to specifics of the DB anyway (Oracle counts nulls = empty strings, for example).

I guess the day has come to make use of this feature.

This is not entirely true. Hibernate/JPA has lots to dislike but it does have constructs for different databases and while it’s not a snap to change db’s It’s not even that bad. We did a recent change on a fairly large application and it took us 3 weeks. 1 week dev and 2 week testing.

It's definitely not a myth. Jira supports PostgreSQL, MS SQL, H2, Oracle. I think that Hibernate helps tremendously with that. Sure, you have to take extra care anyway, but the differences between databases are real and there's enough code in Hibernate to smooth those differences.

Jira doesn't use hibernate.

I'm in the same boat. Default was Oracle, but now we have more options. The order of preference per management is Postgres -> SQL -> Oracle. New applications need a deviation to use Oracle.

You don't even have to be dynamically provisioning anything, even just standard VMware virtualization is not approved for running Oracle DB. They'll audit you and charge you for every physical CPU core connected in your vSphere cluster. It's total bullshit and their big customers are getting sick of it.

Oracle audits are basically just sales tactics to get you to buy their cloud services so they can turn around and pretend they're a leader in the cloud.

It's also a push to get you to buy Exadata/Exalytics appliances, since the lock-in is harder, and margins are bigger.

How do they enforce a audit? What is to stop a company preventing them access?

It's generally built into the contract. Don't let them audit? Lose all your licenses.

A bunch of their stuff phones home too.

They did get a nice short-term windfall by changing the JDK license. Which will fall away in a couple of years as everyone shifts to OpenJDK or something other than Java.

I'm also seeing big stodgy companies getting more comfortable with things like Postgres. It will take a while, but barring some big change, Oracle is headed down long term.

Oracle is not a tech driven firm. It is sales driven, meaning their sales teams are world class at ingratiating themselves with decision makers at "big stodgy companies". How do you explain why they survived all this time. Very very few of the top Tech firms use any Oracle software for atleast the last 10-12 years now. Some even have internal procurement bans!

Sure, not disputing how they got where they are. I'm speculating that none of that helps them from here. Big stodgy companies, their bread and butter, are now tired of Oracle's shit. And, they now have lots of credible options.

They USED to have world class software. Compare Oracle DB to the competitors 10 years ago. They were at least GOOD. Lots of good features for giant companies.

Competitors have gotten a lot better over the last 10 years. Oracle has mostly just raised pricing. It works for a while, but at the current path I don't see Oracle the company existing in 20 years.

I would say arguably they ARE still world class. It is just most people don't need or cant justify their price tag when competition has caught on with clear Roadmap and sustainability.

So basically they have a shrinking value proposition and they are rising prices. ( Why does that sound so much like Apple, Now I am worried )

I love my earpods and I think it shows they still are at the top when it comes to innovating on consumer devices.

Using Apple Pay also makes so much sense for the consumer as it limits your CC data being at every retailer since Apple Pay tokenizes the sensitive information.

If you want to see where Apple is going, checkout the Apple Healthkit API:

https://developer.apple.com/documentation/healthkit/samples/...

While it is too soon to say that Apple will own the market, Apple is the only big tech company that has positioned itself for privacy and trust, giving it the best chance to own this market of the big four.

“Good” depends on the use case. I used Oracle once, about ten years ago. The client library was so big and bloated that just loading it took the majority of the total running time of my not-terribly-long-running program.

Not to mention that figuring out the connection strings is so miserably complicated that you basically need a trained expert to figure it out. The docs were worthless.

I think I can spin up an entire blank MySQL instance faster than I could connect to an existing Oracle database back then.

Well if you don’t need the features then clearly you are not the target. The oracle DB is going to support many different things compared to your MySQL cluster.

I was most definitely not the target. But having such a miserable experience just accomplishing the basics did not make me at all inclined to consider Oracle for future projects.

How are they doing their financials?

>>> Very very few of the top Tech firms use any Oracle software for atleast the last 10-12 years now.

I think you missed the word "new" in there. Amazon finally is cutting ties with its Oracle DB this year. While I won't be surprised if AWS starts offering "migrate off Oracle" as a consulting service, Oracle was still the standard through the 2000s. I think they have built up so much negative goodwill though that no startup would ever consider using Oracle even if they gave the first five years away for free.

> While I won't be surprised if AWS starts offering "migrate off Oracle" as a consulting service

If AWS can offer a database that satisfies the Mongo API (DocumentDB), then why couldn't Amazon offer a database that satisfies the Oracle DB API, complete with automatic migration?

The day that happens is the day Oracle stock goes straight in the toilet.

I suspect that would resurrect something like the Oracle vs Google lawsuit, where an API signature is somehow a thing you can copyright.

Mongo was open source. They copied the api under that license.

My understanding, and I’m not a lawyer so don’t consider this legal advice, is that you can’t copy a commercial api or it’s infringing on their intellectual property.

The mongo example is a bit odd since the open source license was AGPL. Sounds like you can copy an AGPL API signature without releasing your own source?

The copyrightability of APIs (specifically Oracle’s APIs, even!) is at the heart of a lengthy and still-ongoing legal battle, with several decisions in both directions.

I thought the copyright law had an exemption for compatibility?

See https://www.eff.org/cases/oracle-v-google

Oracle won the last round, Google asked the Supreme court to look at it in January. Guess we'll know some time soon.

The "API" is doable. Being binary compatible is not at all. Oracle DB is complex because of historical choices (the patch picking, compatibility flags) etc... inbuilt JVM and PL/SQL.

Those things that make working on OracleDB today difficult is what made it so dependable for people with money in the last 20 years.

It also runs on a lot of architectures that today's software is not really running on anymore.

It would probably turn into the ultimate cigar butt stock though as long as they didn't keep spending money on crazy acquisitions and the stock fell 40% on the news.

I would have to think so many organizations have random scripts, etc. that the risk vs. reward still won still be high er than you think, especially when say the CTO of JPMorgan's job depends on "knowing Oracle."

Personally, I think IBM will likely crash and burn way before Oracle does.

The entire Amazon retail business is on Oracle DB.

It was migrated away to aurora

I don't understand their java strategy at all. By making the switch that they did they forced into existence Amazon Corretto. There's are lots of companies that, rightly or wrongly, want a big corporate name behind a piece of technology they are relying on. IBM and Red Hat were already there, but no CTO is going to have any objections to relying on Amazon over Oracle.

It also may be the last kicker needed for some to start moving to Node, Rust, Go, etc.

Maybe, but I doubt it.

Java is in a rather secure place right now. Everyone that chases the new shiny is long gone but it’s not yet at the point where it is difficult to hire for (e.g. COBOL).

So young buck comes to the CTO and says 'let's do a rewrite in Go' and the CTO says 'why in the world would I want to do that?' Then young buck says 'okay, what about just coding the new frobinator in Go?' CTO still says no, because then he’d have to hire Go programmers and Java programmers or at least waste time cross training. Young buck's less honest compadres have probably gone ahead and put Go into production here or there without permission, but I don’t think that’s a danger to Java.

What really changes this dynamic is when schools stop teaching Java.

Go is legit and a huge threat to java and other languages. My entire career has mostly been java based for the backend. The last few months all I’ve been writing is go. I’m finding libraries in Go that I can’t find in Java. I’m seeing lots of new tech all written in go. Go has all the momentum at this point.

You could edit JVM bytecode by hand, but nobody wants to, we have tools for that. I'm sorely tempted to make a language that transpiles to/from all the Go boilerplate that shouldn't concern human beings.

Can you give me examples of what you mean? I agree some of it is low level, verbose even, but on the other hand, I'm able to figure things out on new projects rather easy, there's just fewer layers and magic things happening.

Some of the boilerplate is likely to go away in new versions. (See new "check" syntax.) But the hidden advantage of Go has always been the ability to look at a function and just see what it does. That comes at the expense of boilerplate, but is worth it.

Whenever I look at some Go, it seems the "what it does" is buried in a big pile of mechanics that could have been generated. You're more likely to see half a dozen error-prone lines copying between slices than an obvious HOF named "filter".

Wouldn't Kotlin or Dart be a better replacement for Java?

Go isn't just a replacement for Java, it's a step off the entire JVM and out of the ecosystem of heavyweight "enterprise" frameworks.

Go is heavyweight: its runtime doubles your memory usage and gives you an incredibly slow garbage collector, it also gets fast compile times partly by not optimising much. Source-wise it has a boilerplate problem practically as bad as Java's just in different ways.

Meanwhile there are lots of frameworks in the JVM world, some of them "heavy" with features, but if you want frameworks that are much smaller, github is overflowing with them.

Kotlin eliminates most of the boilerplate, leaving you with a slimline language, a powerful runtime and a wide choice of libraries. Go doesn't have much to compete with that.

My feeling is that they'll exist alongside each other in not very overlapping niches.

Java is what you write in, if what you're selling isn't the code but some service that the code provides, and you aren't a startup, and you can afford a support team and want a lot of devs.

Go is what you write in when you're selling the thing you made directly, like say it's a DB engine, or you're a startup and want low effort support and only want a few devs.

I imagine java is what a company would choose if they have a massive budget, but is unwilling to hire cowboy coders.

Java makes maintaining a feature easy, and the programmers hired can be average and still be able to make something of acceptable quality, in an acceptable timeframe.

A startup rather wants to move fast. Java is harder to move fast at - but Go probably has that in spades. I do believe Go will not succeed in a large, beaurocratic place. And as a start up grow out to become more and more beaurocratic, the tech transitions with it.

Don't they teach Python now?

Based on my a limited view, if a new grad majored in something other than computer science+ but took some programming classes he or she is overwhelmingly likely to be most comfortable in python. However new grads with a cs degree+ are likely to have at least some and often quite a bit of exposure to java.

+ or software engineering, etc.

What is the practical difference between Oracle Java SE and OpenJDK? That is, why would you pay for it instead of immediately switching to OpenJDK?

They (mostly) didn't used to charge for their JDK. Up until recently, there were code differences. As of v11, there's supposedly no difference.

So, basically, companies need to move from Oracle JDK 8 to OpenJDK 11. And pay Oracle until they are done. It's a fair amount of testing and verification for some big companies. Now that it's costing them license money, they'll do it.

Edit: And, of course, many will also be incented to move off Java altogether.

Support contract in order to have a lightning rod for blame if/when shit hits the fan.

There are a few odds and ends—IIRC some fonts, some gpu accelerated widget implementations for swing, and some obscure crypto options.

But the OpenJDK is the reference implementation and almost certainly has everything you need. The only reason to use OracleJDK is if you want to buy a support contract from oracle (NB there are companies, including very big ones, that will sell you a support contract for OpenJDK.)

There are almost no visible differences between Oracle and Open JDK now. Here's list of differences: https://jaxenter.com/oracle-jdk-builds-openjdk-builds-differ... basically it's some legacy support and their commercial offerings.

You would pay for it if you need support from the very creators of that software. Otherwise you would use free version. It did not change conceptually. All they did is made their offering more visible (and probably they will blackmail those who download stuff without thinking and reading licenses). You could buy support and use proprietary software before that change.

It's pretty obvious that Oracle has been milking the government for decades. Those contracts and license fees that they get are just free money from taxpayers. Does any software developer still use Oracle products? Even Java started to lose momentum after Oracle acquired it.

>Even Java started to lose momentum after Oracle acquired it.

I sometimes wonder how much would it take to buy back Java from Oracle. Basically to buy back the Trademark, TCK and the official stewardship. ( Although I have to admit they are doing a pretty good job so far with Java )

I wish this would be done at some point. So Java could be popular again somehow...

Problem is Java is open source , and companies like Google , Amazon , Redhat have their own JVM implementations... So it’s not so much an issue.

In fact it’s the opposite , having to acquire something as big as Java comes with huge responsabilities to maintain it and make it evolve.

Hence these days , Rust and Golang really got popular , it wouldn’t make sense for anyone to buy back Java...

Android's java ius not a real java implementation.

Use kotlin instead!

Uh, I would argue Java had already lost momentum before the acquisition. Java 6 was released in 2006 and then stagnated; Sun had pushed a new version every two years since 1998, but 2008 came and went without a new Java.

Oracle bought Sun in the first half of 2009, completing in 2010. They actually moved to release Java 7 pretty quickly after that, there was a flurry of activity - mostly meant to reassure the ecosystem about their commitment to the platform, I think. After Java 8, new additions like lambdas seemed to have somewhat rebooted the scene. The ecosystem looks more lively now than 10 years ago, in many ways. It’s definitely a better language to work with, it feels rather less “enterprise” than before.

The move to monetize the JDK is annoying but almost understandable as a compromise between the renewed push to evolve the JDK (with yearly releases) and Oracle’s inclination to spend as little as possible on support. They don’t care if some of that support activity ends up generating money for others (via alternative JDKs), they just don’t want to be on the hook for it.

I’m hardly a fan of Oracle (the cloud move is such a fraud), but to be honest I think their Java strategy does make sense. It certainly makes more sense than Sun’s.

I'm still enraged over the fact that Oracle got $300 million to make a website for Oregon's healthcare and failed. The website wasn't ready (nor would it ever be ready), and the whole state's healthcare was in tatters when the ACA came into effect. Oregon had to scramble to hire thousands of temporary staff and start processing paper forms to keep things running.

Oregon sued and ended up settling ostensibly for $100 million, but in actuality, it was $60 million in "customer service" and $25 million in legal fees.

Obviously, Oregon's government is culpable as well, but it's another data point to suggest Oracle is a dodgy sales/legal company posing as an engineering company.

I wouldn't be surprised if some of the government officials who made this deal possible are now working for Oracle as highly paid consultants.

Reply on news.ycombinator.com