“It's easier to port a shell than a shell script” (1998)

http://gopher.quux.org:70/Humor%20and%20Fun/Microsoft_KSH.txt%7C/MBOX-MESSAGE/1

by beshrkayali • 11 years ago

92 60 11 years ago

gopher.quux.org

readme• 11 years ago

And then microsoft's feelings were hurt so badly that one day they decided to write powershell.

_Ogre• 11 years ago

Somehow I doubt it took them more than 8 years to write Powershell.

steve-howard• 11 years ago

Maybe not, but it does take me 8 hours to get a Powershell script right.

nickpsecurity• 11 years ago

I wish I saw that. That would've been epic.

rawdisk• 11 years ago

I thought this was quite amusing when I read it back in the late 90's. At one point their proposed shell was called "Monad". I wondered "Why do they want to do this?"

I have gone to a number of conferences with MS employees over the years, especially MS lawyers, and with respect to software, they often seem like they have been indoctrinated into a cult. As if the existence of software outside of MS Windows does not exist and as if MS holds the key to solving all problems. But MS does not offer solutions. They create problems (e.g., for anyone who has to deal with their licensing) and then offer "solutions" to the problems they created. It is like someone who proposes to "simplify" something by introducing more complexity.

The idea of MS embracing a proper shell seemed laughable to me. Windows users expect pointing and clicking (and today they want to be able to touch the screen).

Now fast forward to 2015 and it seems like PowerShell has a significant number of users. I am sure it is helping Windows users greatly. Good for them.

But I still think it is laughable. The name still makes me laugh.

I'd rather use something like Back Orifice (which has a simple UNIX client) to control Windows from a proper shell on a computer running UNIX. Or SSH via Cygwin. Or even SUA which has ksh and tcsh. Alas no simple Bourne shell (better for writing portable scripts).

malkia• 11 years ago

Rumor has it, that a somewhat big bank, somewhere in Europe used batch files to drive it's business back in the day...

Natsu• 11 years ago

There was once a time when I as a junior employee was left to create a backup system out of shared drives on the Novell server, XCOPY and AUTOEXEC.BAT.

It worked better than one might expect. You'd get emails about failed backups (backup flag files removed after a successful backup), backups were versioned, and restore was quick & easy (execute a simple script from the shared drive, then sys c: from a boot disk).

dkarapetyan2• 11 years ago

You should try powershell. It's actually pretty good and is much better than unix shells.

KnightHawk3• 11 years ago

It's petty but I can't get over the case sensitive command names with capital letters in it.

They are honestly too long.

vezzy-fnord4• 11 years ago

It might be better than Bourne or C-derived shells, but I have a hard time conceding it beats for instance Inferno shell, es or rc.

viraptor• 11 years ago

Better than Unix tools may be debatable. (Syntax for basic file operations is a bit verbose) But it's definitely better integrated into overall system administration than any shell available on Unix.

kbenson4• 11 years ago

It has it's pros and cons, but in the end it's fundamentally different. Powershell wouldn't work nearly as well on a UNIX system, where much of the tooling is already just text, provided by many different sources, and uses a much more diverse set of utilities. Text makes a good lowest common denominator in that situation.

pjc50• 11 years ago

It's almost better. It's good for automating some system administration tasks. It's not so good for text processing; there are a number of traps and mis-steps. The object pipelines are great, but have to be held in memory (you can't pipe an intermediate state to disk). By default text output wraps at the terminal width even if you're writing to a file, which is just stupid and inconvenient for further automatic processing.

In order to write a script, you first have to change your system security settings.

More subtly, because it's comparatively new it's not very dogfooded and there isn't a rich source of examples. UNIX systems usually have lots of shell scripts in /etc for you to learn from.

x1024• 11 years ago

Haha, yes. Microsoft has always delivered things that are "better" than the standards they are written against.

fresleven• 11 years ago

Seeing a link to a gopher:// site and discovering that Safari in iOS still supports it made my day.

blahedo• 11 years ago

Alas, it's not a gopher:// link; gopher is just the name of the host. (Perhaps it also is a gopher server, or once was.)

lindner• 11 years ago

It's running pygopherd

http://gopher.quux.org:70/devel/gopher/pygopherd

brisance• 11 years ago

I might be missing something; isn't the machine named gopher but it is using HTTP to access it on port 70?

ChristianBundy• 11 years ago

This is correct.

beshrkayali3• 11 years ago

Unfortunately none of the current latest versions of top browse supports gopher afaik. Firefox dropped it at 3.6 (but still possible with OverbiteFF).

Edit: here's the gopher link for it gopher://gopher.quux.org/0/Humor%20and%20Fun/Microsoft_KSH.txt|/MBOX-MESSAGE/1

S4M• 11 years ago

1998

vezzy-fnord4• 11 years ago

Also covered by WikiWikiWeb: http://c2.com/cgi/wiki?KornShellStory

digi_owl2• 11 years ago

So the title of this link is from the title of the email. That in turn is a quote from Larry Wall, who do not seem to be present in the story at all.

beshrkayali3• 11 years ago

Well, I assumed that the point of the quote is that it's fitting for the story, mentioning Larry was for attribution (which I see got removed from the title, I don't mind but I think it should've stayed in there for credit)

digi_owl2• 11 years ago

Yeah, i just came expecting to see something about the issues of porting scripts between shells or something. especially as there has been a whole lot of pro/con shell scripts as of late.

kazinator5• 11 years ago

I could have easily put Korn in his place if I had been that Microsoft person.

Ah, but no; I would end up saying some things that would have been too ironic to come from a Microsoft person, such as:

- why are you Unix people making a POSIX standard which covers the shell command language, if you're going to insist on one reference implementation being used on everyone? Just dump the source code listing of that into a document, add a few comments here and there, and call it the standard.

- isn't it healthier to have multiple implementations rather than a "monoculture"? Do we all use the same C compiler?

- the MKS shell could be made to diagnose and reject the use of Korn shell extensions, helping the programmer write a more portable script.

- wait, did you say your name was Korn, or Bourne? I didn't quite hear you from the back. If you're Mr. Korn, I would appreciate it if Mr. Bourne would now stand up, if present, and ask you why you don't just use the shell with his name on it.

Sir_Cmpwn2• 11 years ago

why are you Unix people making a POSIX standard which covers the shell command language, if you're going to insist on one reference implementation being used on everyone? Just dump the source code listing of that into a document, add a few comments here and there, and call it the standard.

Any shell that isn't compatible with the _reference_ implementation of that shell is flawed.

kazinator5• 11 years ago

The POSIX standard that came into effect in 1990 doesn't contain or assert any reference implementation.

A reference implementation accompanying a document can be at odds with the specification. If a document says that a reference implementation shall be held correct if there is any discrepancy, then specification is greatly diminished in value. It's obviously just a best effort to describe the reference implementation, and not a forceful standard.

What if a buffer overflow is discovered in the reference implementation? Everyone must have the same one to be compatible?

vezzy-fnord4• 11 years ago

You think these are zingers, but they're really just worthy of cringing.

a) Red herring. No one is talking about a reference implementation of Unix shells, but about Korn shell compatibility in particular.

b) Red herring. Don't imply it's a Korn or Korn-compatible shell - that's the issue here.

c) Red herring. Though the conversation was never recorded, it's implied that nothing was said about "portability", per se, but in particular compatibility with the Korn shell language specification.

d) Red herring and an insult. Stephen Bourne didn't write the first Unix shell. Ken Thompson did. Is your implication that no one should iterate or create new designs? Microsoft creating a Unix shell is fine, but the issue at hand is that it wasn't a Korn shell like they claimed it to be.

kazinator5• 11 years ago

Microsoft wasn't creating anything; they were working with MKS:

https://en.wikipedia.org/wiki/MKS_Inc

MKS claimed to have a Korn-shell workalike, AFAIR.

The MPM asserted again that the shell was pretty compatible and should work quite well

which sounds reasonable, and isn't an assertion that MKS has the Korn shell in it such that it is 100% compatible. (And could anything even be 100% compatible, if not running on Unix? The Korn shell source ported to DOS woudln't be 100% compatible with the Korn shell, just "pretty compatible" so that many things work "quite well".)

They probably had some good reasons to with MKS, because in fact it isn't easy to port a shell: not to the DOS environment. Or Windows. The minimal example of doing this reasonably well might today be the MSYS environment used by MinGW, which has to re-create significant infrastructure (a whole C library from the ground up and other things).

Maybe Microsoft also had favorable licensing with MKS, avoiding the AT&T license.

“It's easier to port a shell than a shell script” (1998)

Related Stories