Terms - Proposals

[Ed Shaya]

Roy Williams wrote:

>
>> So a mirror will generally always return the same results as the 
>> original. A
>> snapshot can only guarantee to return the same results up until the next
>> update of the original. An extract can never guarantee to return the 
>> same
>> results.
>
>
> We have astronomical datasets, called A and B.
>
> A delivers a lossless compression of B. Are they mirrors?

. If the dataset is held as compressed and service B automatically 
decompresses as requests come in so that one gets exact copies of A, 
then for all intents and purposes the dataset at datacenterB, with 
automatic decompressionServiceB, holds datasetB and it is a mirror of 
datasetA.   If it requires an extra command in the interface to 
decompress then it is not identical because it is not treated 
identically to get identical results.

>
> When I query A and B, I get a table of results. A and B have different 
> default sorting order. Are they mirrors?

Your examples point out that the term mirror is not applicable to 
services but only to the data.
 Perhaps the dataset is a mirror but the service is different.  Maybe I 
can use serviceA on mirrorB and get same sorting order.  Maybe not.

>
> A delivers much lower bandwidth than B. Are they mirrors?

Sounds like you want to extend mirror to cover services.  That is a 
different issue.  Services are almost never exact copies.  I doubt that 
it is a useful term.  I think we just say, two services are similar if 
for all tested inputs it gave the same output, bit for bit.

>
> A delivers data with calibration version 3.2 and B delivers 
> calibration versions 3.1 and 3.2. Are they mirrors?

Related services, with mirrored datasets, but different parameter 
settings.  It might become a similar service if the calibration version 
is changed to selectable.

>
> A can deliver data from Virgo AND Orion, but B only covers Orion. Are 
> they mirrors?
>
B has a mirror of the Orion data.

> A is available all the time, but B is located in Utah and cannot 
> deliver data on Sundays. Are they mirrors?
>
We agree, mirror is not a good term to use on services.  Services are 
too dependent on system parameters to ever be exact copies.

> A is the server that is mentioned in the peer-reviewed publication by 
> Professor Bigshot. The B data is a copy made by an irresponsible 
> student, and is not mentionsed in the Bigshot paper. Are they mirrors?
>
An illegal mirror perhaps.

Ed