[bug#61789] ‘tor-hidden-service’ deprecation

[Bruno Victal]

Hi Ludo’,

On 2023-03-03 16:43, Ludovic Courtès wrote:
> Bruno Victal <mirai@makinata.eu> skribis:
> 
>> Due to (now renamed) 'hidden-service' record type not being exported,
>> the only way hidden services could have worked is through the now deprecated
>> 'tor-hidden-service' procedure which also had the issue that it can only
>> be used once since the returned service always had the same name.
>>
>> This commit fixes the Tor service documentation and
>> deprecates 'tor-hidden-service' procedure, correcting some inconsistencies 
>> along the way.
> 
> I’m also leaving this one out for now.
> 
> Could you please make it a separate patch series, with each aspect in
> its own patch?  I realize I’m asking you for extra boring work, but this
> should help clarify the kind of changes we’re talking about.

I think it's difficult to split this one into meaningful patches, reason being 
that
'tor-hidden-service-type' can't be used alone since the record constructor for a
Tor hidden service (hidden-service, which is IMO a "collision prone" name) is 
not exported.

The fact that it isn't exported also means that the 'hidden-services field from 
<tor-configuration>
was impossible to configure. Extending 'tor-service-type' was also impossible 
save for the
(to be deprecated) 'tor-hidden-service' procedure which provisions a 
'tor-hidden-service-type'
that is simply a service extension for 'tor-service-type'.

The 'tor-hidden-service' and 'tor-hidden-service-type' are extremely misleading 
to what will
happen behind the scenes should more than one hidden-service be provisioned 
(with 'tor-hidden-service').
Since it does so via a 'tor-hidden-service-type' which has its own name, only 
one of the hidden-services
will get configured, which one = dice roll.

IMO this 'tor-hidden-service-type' shouldn't exist at all and 
tor-hidden-service can safely be
converted into a simple-service that extends tor-service-type.


Cheers,
Bruno