Rugby-core syntax
[1/5] from: rebol665:ifrance at: 10-Feb-2002 11:55
Hi Petr
Thanks for warming me up. I feel better. I like Maarten's idea to build a
simplified rugby. I spend some hours testing the new rugby-core and updating
my notes and you can imagine I was at least surprised by Maarten answer
especially after his ...
>I invite you all to use it and test it, if you find bugs, report them to
me.
>I'll fix it asap.
I think simplicity and consistency are important in software.
You were the first to point that
serve/witch [now] tcp://:9002
is not consistent with
do get-rugby-service http://localhost:9002
And as rugby is now only HTTP, it seems to me that
serve/witch [now] tcp://:9002
can be simplified like this
serve/with [now] 9002
or like this
serve/port [now] 9002
However Maartens own rugby. It's his baby. One he can be proud of . I am
confident he will let us know if he wants some more testing and suggestions,
or maybe just testing.
Patrick
[2/5] from: petr:krenzelok:trz:cz at: 11-Feb-2002 6:09
Hi,
pat665 wrote:
> However Maartens own rugby. It's his baby. One he can be proud of . I am
> confident he will let us know if he wants some more testing and suggestions,
> or maybe just testing.
>
One thing I fear is - someone will suggest, that Rugby is BSD licence, so if we
don't like something, we can produce our own version, e.g. Rugby+. But - if
changes are not going to be made, I will do it at least for my private usage - as
an former amigan, I hate community splits :-)
btw: could you do a following test for me, please? (as according to my
measurements Rugby is now 20 - 30% slower, and I don't know why ...)
console1:
do %rugby.r
serve/with [echo] tcp://:8005
console2:
do %rugby.r
do get-rugby-service http://localhost:8005
start: now/time/precise loop 100 [echo "test"] print now/time/precise - start
... and post the result, especially in comparison to older (4.3) Rugby version?
Thanks a lot,
-pekr-
[3/5] from: koopmans:itr:ing:nl at: 11-Feb-2002 9:13
Some clarifications:
- the tcp://:port-no is close to REBOLs syntax for opening a server port.
This is especially important when you have multiple network interfaces in a
machine, so that you can do tcp://:networkif-of-your-choice:portnumber
So that's why it won't change.
- No security refinements. Currently the security framework is completely
independant of the real rugby core (rexec). It builds on top of that. The
only thing in rugby that knows about sexec is the part tha builds stubs, but
that may change as well. The upside of this that you can remove touchdown_*
from the modules and the build script, and you have a smaller Rugby that
still works. The downside is that you don't have /secure as a refinement but
some extra functions (3).
That is not to say that you may not add your own wrappers and submit them to
me ;-)
- Rugby seems somewhat slower, but that is not the compression (although that
may be useless in some cases) but the http part. OTOH, http provides complete
transparency by default, without any pain if you have a REBOL supported
proxy/firewall. If I make all off these things configurable you end up with a
Websphere, and that's *not* what I want.
And yes, Rugby is my software, but you are free to clone and abuse. That's
why the BSD license is there.
--Maarten
[4/5] from: petr::krenzelok::trz::cz at: 11-Feb-2002 11:16
Maarten wrote:
> Some clarifications:
>
> - the tcp://:port-no is close to REBOLs syntax for opening a server port.
> This is especially important when you have multiple network interfaces in a
> machine, so that you can do tcp://:networkif-of-your-choice:portnumber
> So that's why it won't change.
But ... :-)) No, really, just kidding. OTOH - take it from user's perspective -
what other network interface are you talking about here? Maybe it is just me not
understanding something. If you open tcp://:9005 listen port, you can't open any
http://:9005 or xml-rpc://:9005 on the same one. Or what do you mean by multiple
network interfaces here? And after all - as you said - it is transparent, so
stating serve/with-port [echo] 9005 would be clean too ...
> - No security refinements. Currently the security framework is completely
> independant of the real rugby core (rexec). It builds on top of that. The
<<quoted lines omitted: 3>>
> still works. The downside is that you don't have /secure as a refinement but
> some extra functions (3).
well, it is still solvable, isn't it? what about following:
rexec: func [params /secure][
if secure call-sexec-with-func-args ....
]
> That is not to say that you may not add your own wrappers and submit them to
> me ;-)
>
> - Rugby seems somewhat slower, but that is not the compression (although that
> may be useless in some cases)
OK, is every Rugby packet compressed or just Rugby headers? I think that it
makes sense to compress stubs (textual info), but not data, because if you will
compress some already compressed data (mp3. jpg, other), you can grow in size and
'compress itself will surely add to total communication time. I think that
compression of data can be left to user level functions ...
btw I tried to remove 'compress and 'decompress from 'compose and 'decompose
functions, but was getting an error. I wanted to measure a difference ...
> but the http part. OTOH, http provides complete
> transparency by default, without any pain if you have a REBOL supported
> proxy/firewall. If I make all off these things configurable you end up with a
> Websphere, and that's *not* what I want.
Isn't Webshpere a professional product, done by professionals, sold by
professional company? :-)
> And yes, Rugby is my software, but you are free to clone and abuse. That's
> why the BSD license is there.
I see no point in separate release. Rugby is young. If I am not comfort with
something, I will let you know, along with my reasoning, examples, etc.
PS: I looked at mobile code, but it does not seem to be modified to work with
latest rugby, or does it?
[5/5] from: koopmans:itr:ing:nl at: 11-Feb-2002 13:24
> > Some clarifications:
> >
<<quoted lines omitted: 9>>
> - as you said - it is transparent, so stating serve/with-port [echo] 9005
> would be clean too ...
I mean two different assigned IP addresses on the same machine, a situation
commonly encountered when dealing with multiple subnets.
> > - No security refinements. Currently the security framework is completely
> > independant of the real rugby core (rexec). It builds on top of that. The
<<quoted lines omitted: 10>>
> > to me ;-)
> >
As I said: you are free to do so. It is solvable, I have just reasons (good
ones IMHO) to do it differently.
> > - Rugby seems somewhat slower, but that is not the compression (although
> > that may be useless in some cases)
<<quoted lines omitted: 4>>
> time. I think that compression of data can be left to user level functions
> ...
All rugby data is compressed.
> btw I tried to remove 'compress and 'decompress from 'compose and
> 'decompose functions, but was getting an error. I wanted to measure a
> difference ...
Then you'd need to do the same in tunnel.r (the client).
> > but the http part. OTOH, http provides complete
> > transparency by default, without any pain if you have a REBOL supported
> > proxy/firewall. If I make all off these things configurable you end up
> > with a Websphere, and that's *not* what I want.
>
> Isn't Webshpere a professional product, done by professionals, sold by
> professional company? :-)
>
Exactly. They are outnumbered, at least qualitatively ;-)
> > And yes, Rugby is my software, but you are free to clone and abuse.
> > That's why the BSD license is there.
>
> I see no point in separate release. Rugby is young. If I am not comfort
> with something, I will let you know, along with my reasoning, examples,
> etc.
>
I know, and your feedback *is* appreciated. But some things are a matter of
taste, which is just that.
> PS: I looked at mobile code, but it does not seem to be modified to work
> with latest rugby, or does it?
>
Correct. I didn't think anybody would find it useful.
--Maarten
Notes
- Quoted lines have been omitted from some messages.
View the message alone to see the lines that have been omitted