[9fans] Re: [inferno-list] Shoot

[9fans] Re: [inferno-list] Shoot

[Uriel]

(CCing 9fans so everyone can share the fun)

On Sun, May 22, 2005 at 07:42:59AM -0700, Jack Johnson wrote:
> TiddlyWiki is 130k of code and content, that goes a long way in Limbo
> (though it discounts the overhead of the VM and supporting software).
That also discounts the millions of lines of crud those 130k sit on top
of, a few times more code than the whole Plan 9 and Inferno systems put
together.  

Why there are no web browsers for Plan 9? Why Charon doesn't handle all
that stuff? Because it would take many years of hacking and masochism,
and then you would end up with a lousy result, because the web is too
inconsistent.

> TiddlyWiki concept of a serverless, single-document wiki is a powerful
> one, and working with it for extended periods is enlightening; it's
> like using a word processor without the word processor.  
You mean it's like... using a text editor on your local file system? Now
that is what I call progress!

Only that when you use a text editor you have the /bin/* toolset in your
fingertips; with TiddlyWiki you have to use the hideously awkward text
input field on your clunky web browser... I love it.

How does the wc(1) of the web look like? Yet another plug-in with yet
another context menu?

> It makes me hope Google is working on an OS, because I want more of
> this.
I'm seriously disturbed by the direction google is taking and how they
are locking out everyone that wont buy into the web browser scam.

> After having just come from a workshop at Microsoft on OneNote and
> thinking about the (non-portable) baggage under that hood, the
> standards soup that Russ rattled off seems like cake.
"Please note that this site works best in Firefox[1]. It will also work
almost as well in Internet Explorer[6]. It will not work properly in
Opera or Safari" You call this portable? Firefox is >6 million lines of
C++ that will run on windows and X/POSIX.

> Can it be simplified? Sure. Where would you start?
I think that you, like many, are very confused; the success of 'web
applications' has only one reason: easy deployment. People working for
Google have said many times: it's great to deploy a bug fix to hundreds
of millions of users in minutes.

But that is just a hack, deployment of applications on Unix and Windows
was so painful the people tried to find any way around it.

The solution to that problem is not completely clear to me, but I think
Plan 9/Inferno with network transparent filesystem-applications solve
most of the problem.


A completely unrelated issue is how to do hypertext in a sane way, in my
opinion the main requirements are: global namespace organization, and a
textual interface for such namespace; it seems a good fit for Plan 9,
the land of textual interfaces and namespaces.

Plumbing and 9P solve 70% of the problem, all you need is a syntax for
global namespace identifiers(URL replacement) and and a protocol to
access the resources, perfect task for 9P.

In the end, as Unix and Plan 9 have demonstrated again and again,
textual interfaces are the key, with a clean and uniform representation
where simple tools can work on any inputs and their output can easily
become the input of the next tool.

The web its a throw back to the days before McIlroy invented pipes...
very dark times indeed.

You should read utah2000:
http://www.cs.bell-labs.com/who/rob/utah2000.pdf

uriel

Re: [9fans] Re: [inferno-list] Shoot

[Jack Johnson]

I don't mean to start a flame war, but...

On 5/22/05, Uriel <uriell@binarydream.org> wrote:
> That also discounts the millions of lines of crud those 130k sit on top
> of, a few times more code than the whole Plan 9 and Inferno systems put
> together.

I don't mean to discount that at all, either.  In fact, it was to put
in the same context as the point you made earlier in the conversation:

>On 5/22/05, Uriel <uriell@binarydream.org> wrote:
>> Nah, we got this 6 million lines of crud implementing a mountain of
>> braindamaged 'standards', tons upon tons of inconsistent ad-hoc hacks
>> everyone implements differently... all to get what would take a handful
>> of lines of code to write in rc or Limbo.

Meaning that no matter what system a clean, sane solution rides on top
of, it still comes with baggage.  And I still ask, what does the end
result look like?  I'm doubtful we'll change the installed base of the
rest of the planet, so do we ask everyone to run an Inferno
application to get the same functionality that their browser already
provides?  Most users could care less about those 6 million lines of
code unless it takes more than twenty minutes to download or breaks
and can't be fixed.

Now, sure, maybe it's already broken.  I buy that.  But now your
grandmother or your third grader is going to use its replacement. 
What makes that repacement qualitatively better?

> and then you would end up with a lousy result, because the web is too
> inconsistent.

It was designed for inconsistency.  It's the act of trying to make it
consistent that breaks the design.

> You mean it's like... using a text editor on your local file system? Now
> that is what I call progress!

It's more than text.  It's interacting with the result.

Many of us use editors outside the Plan 9 and Inferno realm that let
us collapse sections of code to give us another view of the same
information.  Sometimes those kinds of interaction prompt us to
consider a new model (or an old one).

After all, that extra window in sam is awfully handy, and vi *is* just
and editor.

> How does the wc(1) of the web look like? Yet another plug-in with yet
> another context menu?

It depends.  What the hell are you doing using wc(1) on the Web?  Now
grep, maybe....

> "Please note that this site works best in Firefox[1]. It will also work
> almost as well in Internet Explorer[6]. It will not work properly in
> Opera or Safari" You call this portable? Firefox is >6 million lines of
> C++ that will run on windows and X/POSIX.

...which is an astounding feat for >6 million lines of anything.  And
could hardly be said of our own tools before the advent of plan9ports.

> Plumbing and 9P solve 70% of the problem, all you need is a syntax for
> global namespace identifiers(URL replacement) and and a protocol to
> access the resources, perfect task for 9P.

What about rich expression of and interaction with the content (beyond
programming)?  Getting data from point A to point B is old hat. 
That's not how we get to 6 million lines of code.

> In the end, as Unix and Plan 9 have demonstrated again and again,
> textual interfaces are the key, with a clean and uniform representation
> where simple tools can work on any inputs and their output can easily
> become the input of the next tool.

For most users on the planet, they don't look at the content as
resource for another tool, but as the end product.  It's about their
relationship to the content that drives their actions.

The separation of user and programmer is one of the reasons we have
these circular debates, and one of the reasons even our best advocates
turn to other platforms for at least part of their daily work.  And
yes, I mean work.

> Or you could use sane plain text formats and awk... or s-exprs and
> scheme, or...

See, now s-expressions and scheme, there's a great, implementable
idea, full of expressive possibility for the end-user.  Exactly the
kind of thing I was seeking, without dragging us both into a diatribe.

-Jack