Comment by userbinator | Hacker News Reader

Comment by userbinator | Hacker News Reader

One of the things you'll quickly learn from working in IT (and probably other fields) is that "not supported" can mean anything from "physically impossible" to "we don't want you to".

In the case of RAM, "physically impossible" would mean something like the CPU not having that many address bits or the PCB traces not routed, but there are a lot of different not-well-advertised configurations of DIMMs available with the same size[1], so it could be that the manufacturer specified a lower maximum just to avoid having to answer subtle questions. For example, 2x8GB may be OK but 1x16GB not.

It could also be a "X if Y except Z else B" situation, and they just couldn't be bothered to document all the possible combinations and/or explain the details of the PC system architecture that result in such limitations.

To test if all the memory is present (and if it all works correctly), running something like MemTest86 might be sufficient.

[1] https://en.wikipedia.org/wiki/Memory_geometry

> "not supported" can mean anything from "physically impossible" to "we don't want you to"

More like:

from "physically impossible" to "we haven't tested that"

> they just couldn't be bothered to document all the possible combinations

Each combination adds exponential testing and documentation requirements.

Not tested != doesn't work. But that doesn't make not supporting it malicious. There are practical and financial cost to testing every combination, ultimately born by the consumer.

My favorite is when they say "not supported" and it's easy to make it work and is just their way of scaring Enterprise customers to their bread & butter. (One example that comes to mind is a bunch of Rigol oscilloscopes that, even though they have FPGA's on them, can still have more functionality enabled with resistors.)

As an owner of a lower-end Rigol oscilloscope, this is very interesting to me...

> "we don't want you to"

Can be:

- Something that will be rather obsolete soon, so it won't be tested

- Something that usually works, but has poorly understood corner cases or other implications

- Something that is so infrequently used that it's not worth it to build up proper testing infrastructure to prevent regressions

In this case, who thinks Foxcon has an engineer on call, who's familiar with the motherboard, who can actually answer this question? I personally doubt it.

Reply on news.ycombinator.com