So for example, a stepping stone for piping might be:

"Find me a bus 19 that leaves from home to arrive at work before 9am every day for the next week."

"Ok, here are the buses that match your requirements."

"Add those to my calendar."

"Ok, I've added them."

"Also remember to perform this task again every Sunday evening."

"Ok, I'll remember."

That's not exactly what you suggested, but it's close enough and not that far from what Siri is already capable of offering in theory, if it was plugged into the right services.

That's no doubt difficult, but you just seem to be implying it can't be done, without giving any reasons why.

Being able to control and input data to an increasing number of applications reliably is undoubtedly very good, but considering the many many many failed attempts to replace text entry to describe anything mildly complex suggests that it's not going anywhere.

No it doesn't. Simply that a lot of people have tried something and failed to achieve it doesn't mean anything. Lots of people failed to make flying machines, until someone did.

You may be able to give particular reasons why it is difficult and why you think it not likely to be achieved, but the fact, by itself, that no one has been able to do it so far, doesn't demonstrate anything.

Yes it does. It doesn't prove anything, but it does mean something.

I'd point out that we've surrounded command lines with a lot of accoutrement over the years, like history, conditionals, and various ways of outputting things such that we can debug our system. The utility of the command line would plunge if we didn't have any of those, and it is completely unclear how to add those to Siri without turning it into simply another programming environment, a voice-activated interface to Yahoo Pipes.

I'd submit that there is a certain irreducible complexity in the command line, and despite how fuzzy and wonderful it may make you feel when using it, wrapping a voice interface around this complexity makes it worse, not better. Or we'd already be doing it. If this is trying to do the mathematically impossible task of reducing a task's complexity below the number of bits it actually has, then yes, it is impossible, too.

No it doesn't. The history of science and technology is filled with things that people couldn't achieve, that were considered impossible, and then were done.

There is no way you can simply look at the fact that it hasn't been done and is considered hard and conclude anything about whether it can be done or not.

To hopefully avoid a potential misunderstanding: you can give specific reasons why X seems hard to do and not likely to be achieved, and these can mean something. But it's these reasons that mean something, not the fact that it hasn't been achieved so far.

I would counter your claims with the simple fact that the earliest command-line interfaces had none of these. A process change only has to offer an incremental improvement over what came before to drive adoption. Those of us who used the earliest shells and the job-control languages that preferred them know that the bar for usefulness can be set quite low and the tool will still be used and improved upon.

If I'd meant "been eliminated", I would have said that. Nevertheless, I cite the "your mom" argument. "People" are terrified of the nice command line, and we think that they're going to switch to a vastly, vastly more user hostile one, no matter how friendly and chipper it may seem on first glance? If "people" wanted command lines that much they'd already be using them.