Hacker News Reader

Comment by jjice | Hacker News Reader

The absurdity of not being able to support a name as common as Christopher or anything as long or longer just screams "government work". What the hell went through everyone's head when they built this system? Absolutely no testing or real data was used either, but that shouldn't matter, because having a max limit on someone's (very common) name is honestly impressive. The fact they developed this system without addressing this issue is a testament to the quality of government software development.

I'm sure there is good government software out there, but there are plenty of showcases of the opposite (especially since these are systems that NEED to work).

View on HN

So Christopher is an interesting case because it's the longest common English first name at 11 characters. There are names that seem longer but aren't, like Maximilian.

My first son is named Christopher, and we realized right away that there is a 10 character first name limit in a ton of systems almost from day 1 - calling my insurance company, the automated system asked "Are you calling about... Christophe?"... in a French accent, which was hilarious.

It was common in the 1980s for compilers to have arbitrary implementation limits. The C Standard even lists minimums, like 127 nesting levels of blocks, 12 pointer declarators, 4095 characters in a string literal, etc.

My compiler started out with those, but I quickly realized that there were only two actual limits:

1. allocated memory

2. stack size

It turned out to be much less code and many fewer error messages to just detect out of memory and blowing up the stack.

In some programs (compilers included), removing the limits and relying on detecting OOM is not a bad design decision. In others (most notably network-facing programs, but there are other categories), it can be quite a big problem to not have limits on inputs.

The C standard has those limits mostly because you cannot meaningfully define conformance without them.

Limits will be there whether you standardize minima for them or not.

For instance, no C compiler will handle a function with any number of parameters; as you try test cases with more and more parameters, eventually they will all crap out, right? If for no other reason, than memory.

So then if you don't have anything in the standard about this (like the ECMAScript standard doesn't, for instance) you have to conclude that all those compilers are nonconforming.

But if you have it in the standard that a conforming implementation must handle functions with 128 parameters, then those implementations which crap out at over 128 are spared from being declared nonconforming. Only those which crap out at 128 parameters or fewer have a conformance issue.

Limits settle questions. The user reads the spec and then has confidence they can have 4095 characters in a literal which will be portable to anything calling itself conforming. The implementor reads the spec and then has confidence that their implementation doesn't have to handle more than 4095 characters in a literal in order to handle maximally portable programs. There are goalposts firmly in the ground.

It can be useful for an implementation to optionally diagnose when a minimum limit is exceeded, even if it allows for much more. For instance, to tell the user they are using a larger character literal than 4095. This helps the user prepare a portable program that will work in other implementations.

Limits can work, but IIRC the original C would have set the limit on meaningful length of a variable name at 8 (or was it 6?). That... turned out to be a bad place to set that limit. So the limit had to be changed. But the more support the tooling gives for "preserving portability", the more support it gives for making it hard to change the limit when it needs to be changed.

The 6 limit was due to ancient linkers on Unix, which truncated external names to 6 characters. This affected file-scope identifiers with linkage (external or internal).

If a compiler has to work with some object file format which allows only six character symbol names (with case being indistict), there is nothing that can be done; but there is little reason to impose such a stringent limit if the underlying linker technology has a much more lenient one.

I think C90 still made that 6 character the minimum limit; a conforming C implementation of C90 was not required to accept a program which makes two external definitions that do not differ in the first six characters of the name, ignoring case.

In those days, it would have been useful to have a warning if you're using longer identifiers, or if your identifiers are not unique in the first N characters.

C99 raised a bunch of these limits, including that one; it became 31 characters. No mention is made that case is indistinct; it must be distinct.

The limit is still there; use 32 and your program is not strictly conforming to C99.

> What the hell went through everyone's head when they built this system?

The way to make a profit on government contracts:

1) Underbid

2) Find every possible use case that’s OBVIOUSLY needed, yet not in the specs

3) Leave them unimplemented

4) Charge through the nose for the extra work.

The system has already been paid for and put to production. It’s too late for the buyers to back out, lest they be out both money and face

It backfired this time though... the government was like "we can just live with 10 character first names".

>What the hell went through everyone's head when they built this system?

They thought that the next computer system would be better, and when they re-wrote it for the new machine, they'd be able to fix the problems they found, in about 3 years or so. They certainly didn't expect it to still be running in the 1970s or 1980s, let alone in 2021.

IBM broke everything when they introduced backwards compatibility. It saved a ton of time, in the short term, but everything before that point was frozen, and the technical debt it caused has never been paid.

CVS also has trouble with my name (I’m also a Christopher) as do some other private entities like my doctor’s office. So this may not be entirely a government thing.

You can go to a new doctor or pharmacy. The consequences for them not having your full first name recorded are between minimal and nonexistant.

To be a US resident, you must be known to the government. To leave, the government will tax a percentage of your wealth. Try to tell the government to pound sand while remaining, and they can bring the full legal weight of their monopoly on force against you.

The difference in scale of harm between what CVS is capable of versus the government means we should have no issue with holding government to a much higher standard, And be proactive in pointing out when it falls short.

What the hell went through everyone's head when they built this system?

That decision probably dates back decades and was based on the limited amount of space on a 80 column punch card. And as the system evolved past punch cards, no one bothered to update the spec because "it's always been 10 characters, if we change it now, something might break"

This likely happened due to budgetary constraints. Storage space was much more expensive then. 10 extra bytes x 10 million records is 100 megabytes of hard drive space. That's a few thousand dollars in 80's money, just for storage. That doesn't include redundant drives, backups, extra RAM, and whatever else is needed.

The limits have probably been in place for 40+ years and nobody wants change it. Or maybe nobody knows how. It's much easier to tell them to abbreviate their name as "Chris" and move on.

Normally I don't recommend rewriting code, but maintaining such a system is probably more difficult in this case.

> What the hell went through everyone's head when they built this system?

Not sure what did go through, but I'm sure that "Christofer" as a test string didn't.

> just screams "government work"

Or the government not having enough funds to upgrade its systems since the 80s when storage space came at a significant premium (or not making modernizing IT systems a policy/budgetary priority)

... because the same people who scream 'government work' also vote to cut taxes.

Yeah, "... but see how incompetent they are?!!1! Why give them more money?!!1!"

Not to say that throwing money at government solves problems either. A competent administration is needed also to instil confidence in voters that they are actually committed to solving problems. Not sure what the situation is like in NJ

The solution is to support and elect good people. If you don't maintain your house, your house will leak and corrode - no matter who else you blame or what excuses you make.

I agree, but I guess it's hard if there's no good options up for election.. or if they promise one thing and do another. Its particularly the case with winner-takes-all democratic systems. I think representational democracies do better at reflecting the desires of voters, otherwise it's just the incumbent pandering to swing voters

To be completely frank, I hear this from many people and I think they are excuses. I see plenty of successes, including where I live. It's not at all perfect, but a lot can get done. The biggest obstacle is the people standing on the side saying nothing can be done - while other people are doing the work.

Ah, I'm not advocating complacency or inaction, but I agree, things can change even in a flawed democracy, just much slower, and also this can disenfranchise voters. In completely dysfunctional and corrupt government, I think the only option would be to hit the streets en masse, but I don't think NJ is at that point

Sounds like the media player I just bought that errors out with "too many songs to shuffle". And you're hosed.

> screams "government work"

The 'government work' was probably done by a private contractor. Note that the OP is about a private bank. If you haven't seen results like that in private business, you've lived a charmed life!

Wait until you find out about Y2K

Wait until we start to reach 2038, it's going to be either awesome or an epic shitshow

Reply on news.ycombinator.com