Ignoring Linux compatibility would have knock on effects because Linux is now the platform for so many other tool, e.g. Docker.
Regardless of how easy the 'upgrade and migrate to Linux' story is right now, I think it's vital that MS started down this road. Otherwise the .net framework would be [seen as] finished now, and people would start planning to migrate off it.
Now that practically all the features are available in the standard edition (depending on your support and scale-up/scale-out requirements) it could end up used in production for licensing cost reasons particularly if the availability/DR/scaling features work well (where previously if you needed Enterprise edition the cost of windows was a drop in the ocean - Standard isn't exactly pocket money cheap but still...).
Another PoV could just be that these days interpolarity matters much more to MS than it used to, as their key product focus is moving further away from the OS & individual services and more onto the platform with Azure - they don't care what DB engine you run on which OS, as long as you do so on their platform - hence moves to support postgres & mysql in the database-as-a-service-upon-platform-as-a-service manner that they already offer for SQL Server. Of course they'd prefer that you used Windows and SQL Server but not wanting to (or depending on the project, not being able to) use one needn't get in the way of using the other, not should not using one (or either) get I the way of using Azure.
MS SQL is the product that people are choosing, not Windows. Why should the MS SQL team force people to choose Windows?
I'm incredibly happy Microsoft is bringing SQL Server to Linux as a DevOps person, I'm really hoping SSIS/SSAS/SSRS makes it in eventually because they're the only thing keeping me from saying we should be deploying the new cluster on Linux instead of Windows.
Especially since 2016sp1 when they unlocked a lot of previously enterprise edition only features in standard edition. The scalability is still locked down a bit but those limits have crept up over time also: there is a lot that you can do before 128Gb RAM & 24 cores spread between up to 4 CPUs becomes limiting (core engine only limits there, OLAP services, reporting services, and the mem-optimised engine parts have extra limits on SE compared to EE).
Also a chunk of enterprise licensing is in support which if you are truly at that scale might be worth paying for. But if your company/product isn't there yet that is a cost worry to put off to the future, and as developer edition (which feature-wise is enterprise edition, you just can't use it in production) is essentially free you can test in advance to see when/if that expenditure is going to be worthwhile.
Of course it doesn't beat Free on price, so if postgres has the features & support you need, which for many it has, then it is far from an open/shut case. But compared to the other commercial products in the same market...
> it's been an absolute rock
Agreed. I can't think of another product that I've used nearly as much which has caused us so little hassle over the time. It certainly hands MS's other products their arses in that respect.
I don't think this (in the end) is anything more than a marketing trick, there to give the illusion of choice.
This is not uncommon in ISVs, btw. There are many commercial offerings that have a Linux version, crippled in so many interesting ways that it makes it pointless. I'm a Linux person but I advocated against choosing Linux versions (of different products) many times in the past. Actually, that means the marketing trick works.
In my experience, software always has a primary plaform where it shines, and the other "supported" platforms are always stepchildren.