[sword-devel] CrossWire mirroring
Andrew Thule
thulester at gmail.com
Mon Jan 7 12:04:03 MST 2013
Ok. In your example you have two levels of mirrors, root and banch.
If we assume all mirrors are synced (exactly the same), the client (if it
supports more than a preferred mirror) will check its mirrors in the order
they are specified.
So in your case, Y will check A. D. and E. in that order (assuming it
checks all of its configured mirrors)
If all mirrors are exactly the same however, it doesn't need to check A. D.
and E. It only needs to check A. If A is down, it will check D. etc.
If a change is made at the top level however, say an ESV update, there is a
propigation time it takes for the mirrors to catch up. With high volume
Linux mirrors this is minutes. If Crosswire were to have mirrors, it might
have a policy that says something like mirrors are to syncronise no more
than 24 hours no less than 6.
That means in your demonstration it may take A. D. and E. upto 24 hours to
reflect the change. Unlike Linux distributions however, its likely there
can be some tollerance for such propigation effects.
~A
On Mon, Jan 7, 2013 at 1:52 PM, DM Smith <dmsmith at crosswire.org> wrote:
> I guess I need more information on mirrors.
>
> Let's say that there are to mirrors X and Y. For what ever reason, X has
> A, B, C, D and Y has A, D, E. Software is configured to use Y. When it goes
> to get a list of files, what does it get? If it requests B, what does it
> get? Same questions for software configured for X and request for E.
>
> -- DM
>
>
> On Jan 7, 2013, at 1:06 PM, Andrew Thule <thulester at gmail.com> wrote:
>
> My experience with mirrors is that mirrors are done at the level of the
> Operating Systems. Tools like 'rsync', 'lsync', 'chron' etc manage the
> integrity and distribution of these things.
>
> That said, I think what you're saying is that you believe the Sword client
> needs some additional support to support mirrors, and I don't think that's
> entirely true. It may be that support for mirrors would be better
> accomplished through additional patches, but as it stands now I get all of
> Crosswire's modules as a proxy through either my public or private repo for
> the very simply reason it's safer for me to do so.
>
> I have no trouble getting Sword/Crosswire modules from a 'mirror' (as a
> proxy), technically which means most clients (though perhaps not all) can
> manage this.
>
> So I'm speaking specifically about Bibletime, Xiphos, Alkitab,
> PocketSword, and Eloquent. If there are other clients out there, and
> Crosswire pursues 'mirrors', that decision will influence development.
>
> ~A
>
>
> On Mon, Jan 7, 2013 at 12:55 PM, DM Smith <dmsmith at crosswire.org> wrote:
>
>> Mirror management is a moot issue if the software doesn't support
>> mirrors. I have no plans to add such to JSword, unless it is added to SWORD
>> first. I highly doubt that it will be added to SWORD until a problem with
>> resiliency creates a real need. Even then, I'm not sure that that will be
>> used as a solution.
>>
>> In His Service,
>> DM
>>
>> On Jan 7, 2013, at 12:39 PM, Andrew Thule <thulester at gmail.com> wrote:
>>
>> DM, I agree that not having thought through mirror management
>> procedurally (policy and best backpractice) is reason enough to hold off on
>> such a venture, but those problems are typically trivial to solve given
>> effective communication.
>>
>> Since technology is subordinat to intent, what needs to be worked out to
>> move forward along these lines is the clarification of intent.
>>
>> If I were to look back on this recent discussion, I'd suggest a number
>> of principles are already clear:
>> -CrossWire resevers the right to approve or deny mirrors
>> -CrossWire reserves the right to define which repositorories are
>> considered "root" repositories (so authoritive)
>> -Distribution of modules exclusively licensed to CrossWire should be
>> retained by Crosswire
>> -All Mirrors should take no longer than x (period of time) to accurately
>> reflect 'root repositories'
>>
>> Etc.
>> This issues have already (to some extent) been hashed through in debate,
>> however discussion on the matter was limited, unproductive and unclear
>> simply because of the degree of hostility the 'idea' of mirrors alone
>> produced.
>>
>> If there had been an attitude of 'open but undecided' reservation about
>> the matter, rather than outright 'hostility' its likely the issues you are
>> raising now could have been dealt with more readily.
>>
>> ~A
>>
>>
>> On Sun, Jan 6, 2013 at 2:17 PM, DM Smith <dmsmith at crosswire.org> wrote:
>>
>>> A few more reasons we discourage mirroring:
>>> SWORD and JSword have no means for managing mirrors. They expect each
>>> repository to be a unique collection of modules.
>>>
>>> A mirror that is partial, not containing all that is in the master
>>> repository, probably will be confusing to users.
>>>
>>> A mirror that is hosted along with questionable modules probably will
>>> give the appearance that CrossWire condones those modules. Especially when
>>> the modules are in the same repository.
>>>
>>> In Him,
>>> DM
>>>
>>> On Jan 4, 2013, at 9:09 AM, DM Smith <dmsmith at crosswire.org> wrote:
>>>
>>> From time to time, interest has been expressed in mirroring CrossWire's
>>> SWORD modules. I thought I'd reiterate our policy.
>>>
>>> We strongly, very strongly, discourage mirroring of the SWORD module
>>> repository.
>>>
>>> Those modules for which CrossWire has obtained distribution permission
>>> from copyright holders must not be mirrored. These have
>>> "DistributionLicense=Copyrighted; Permission to distribute granted to
>>> CrossWire" in their conf. CrossWire maintains correspondence for each
>>> of these modules.
>>>
>>> Mirrors are seldom current/correct, despite intentions.
>>>
>>> On occasion we have unintentionally hosted modules for which we did not
>>> have permission. When presented with an ownership claim, we typically will
>>> take the module offline immediately and validate the claim. If the claim is
>>> false, we will put the module back up. If the claim is true, we obtain
>>> permission before putting the module back online or we don't put it back.
>>> This is an important part of our stewardship.
>>>
>>> If or when CrossWire has problems with distribution, we'll tackle the
>>> problem at that time. Probably with a "Content Delivery Network" of
>>> CrossWire owned servers.
>>>
>>> This or some variation of this probably should be in the wiki.
>>>
>>> In His Service,
>>> DM
>>>
>>> _______________________________________________
>>> sword-devel mailing list: sword-devel at crosswire.org
>>> http://www.crosswire.org/mailman/listinfo/sword-devel
>>> Instructions to unsubscribe/change your settings at above page
>>>
>>>
>>>
>>> _______________________________________________
>>> sword-devel mailing list: sword-devel at crosswire.org
>>> http://www.crosswire.org/mailman/listinfo/sword-devel
>>> Instructions to unsubscribe/change your settings at above page
>>>
>>
>> _______________________________________________
>> sword-devel mailing list: sword-devel at crosswire.org
>> http://www.crosswire.org/mailman/listinfo/sword-devel
>> Instructions to unsubscribe/change your settings at above page
>>
>>
>>
>> _______________________________________________
>> sword-devel mailing list: sword-devel at crosswire.org
>> http://www.crosswire.org/mailman/listinfo/sword-devel
>> Instructions to unsubscribe/change your settings at above page
>>
>
> _______________________________________________
> sword-devel mailing list: sword-devel at crosswire.org
> http://www.crosswire.org/mailman/listinfo/sword-devel
> Instructions to unsubscribe/change your settings at above page
>
>
>
> _______________________________________________
> sword-devel mailing list: sword-devel at crosswire.org
> http://www.crosswire.org/mailman/listinfo/sword-devel
> Instructions to unsubscribe/change your settings at above page
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.crosswire.org/pipermail/sword-devel/attachments/20130107/09519f09/attachment-0001.html>
More information about the sword-devel
mailing list