R devel - Oct 2009 - Specifying (set of random) port numbers for the help HTML server (Was: Re: Fixing html help port in 2.10)

On Fri, Oct 16, 2009 at 7:54 AM, Duncan Murdoch <murdoch at stats.uwo.ca>
wrote:
> On 10/16/2009 10:39 AM, Dieter Menne wrote:
>>
>> I noted that the new html-help in 2.10 under Windows uses a random port
on
>> my
>> computer.
>> This cause a problem, because when I create a link such as:
>>
>> http://127.0.0.1:28027/library/stats/html/addmargins.html
>>
>> this is for one-time use only. Is it possible to fix the port?
>
> Not currently. ?If you look in tools:::startDynamicHelp you can see how it
> is set up; you could duplicate that setup as a temporary workaround. ?You
> might be better off to build R with static help instead. ?What is the
> application where you want to be able to give out links? ?Perhaps we could
> consult an environment variable or option() to choose the port instead of
> leaving it completely random.
>
> The reason it is random is the worry that multiple R instances on the same
> machine might collide. ?All processes on the machine see the same ports.
Related:
I'd like to suggest an option/environment variable that specifies the
set of port numbers sampled from, e.g. 6800:6850 or similar.

The reason is that in some places, the user don't have admin
privileges on the computer and most/all ports are blocked from running
servers, and there is a sysadm that needs to approve each port to be
unblocked.  With such a security polices it is possible to have the
smaller set of port number unblocked (but not the full 0-65535 range).
 This applies at least to "regular users" on Windows.

My $.02

/Henrik
>
> Duncan Murdoch
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

On 17/10/2009 12:08 AM, Henrik Bengtsson wrote:
> On Fri, Oct 16, 2009 at 7:54 AM, Duncan Murdoch <murdoch at
stats.uwo.ca> wrote:
>> On 10/16/2009 10:39 AM, Dieter Menne wrote:
>>> I noted that the new html-help in 2.10 under Windows uses a random
port on
>>> my
>>> computer.
>>> This cause a problem, because when I create a link such as:
>>>
>>> http://127.0.0.1:28027/library/stats/html/addmargins.html
>>>
>>> this is for one-time use only. Is it possible to fix the port?
>> Not currently.  If you look in tools:::startDynamicHelp you can see how
it
>> is set up; you could duplicate that setup as a temporary workaround. 
You
>> might be better off to build R with static help instead.  What is the
>> application where you want to be able to give out links?  Perhaps we
could
>> consult an environment variable or option() to choose the port instead
of
>> leaving it completely random.
>>
>> The reason it is random is the worry that multiple R instances on the
same
>> machine might collide.  All processes on the machine see the same
ports.
> 
> Related:
> I'd like to suggest an option/environment variable that specifies the
> set of port numbers sampled from, e.g. 6800:6850 or similar.
I've added an option "help.ports" to set this.  It will try your 
specified ports in order; if you want a random selection, permute them 
yourself, e.g.

options(help.ports = sample(6800:6850))

This will make it into 2.10.0.

Duncan Murdoch
> 
> The reason is that in some places, the user don't have admin
> privileges on the computer and most/all ports are blocked from running
> servers, and there is a sysadm that needs to approve each port to be
> unblocked.  With such a security polices it is possible to have the
> smaller set of port number unblocked (but not the full 0-65535 range).
>  This applies at least to "regular users" on Windows.
> 
> My $.02
> 
> /Henrik
> 
>> Duncan Murdoch
>>
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>>
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

Duncan Murdoch-2 wrote:
> 
> 
> options(help.ports = sample(6800:6850))
> 
Nice. So just in case I want

options(help.ports = sample(6800:6800))

What would be the adverse effects (besides me having to check that that port
is reasonable)?

Dieter

-- 
View this message in context:
http://www.nabble.com/Re%3A-Specifying-%28set-of-random%29-port-numbers-for-the-help-HTML-server-%28Was%3A-Re%3A-Fixing-html-help-port-in-2.10%29-tp25935271p25947460.html
Sent from the R devel mailing list archive at Nabble.com.