Subject Re: Record Encoding
Author adem
Jim Starkey wrote:
> adem wrote:
>
> >No, not for each blob *field* , I meant _*each*_ blob.
> >
> >IOW, client tells the server what MimeType *each* blob is and the
> >server treats it as it sees fit.
> >
> No, it doesn't work that way.

I am not sure if it is bad etiquette to rekindle a past
topic, but, I was intrigued if, in general, 'it doesnt
work that way'. Here is a brief exhange on a similar
topic on NexuxDB, you might fnind interesting.

> > In all RDBMS I know a blob is a blob --and that's it.
> >
> > But, it is like saying a number is a number and leaving
> > it at that. In real life, a blob is not a bag of random
> > bytes, it almost always has a mime type asscoiated with
> > it.
> >
> > And, why is this important? Well, it's been long since
> > RDBMSes were simply the keepers of those bytes; one can
> > manipulate a lot of the stuff in them using SPs.
> >
> > Yet, when it comes to blobs, there is no intrinsic way
> > for a SP to know what that blob contains --unless of
> > course you have another field in there telling it what
> > it is, which is a clumsy way of doing it.
> >
> > OK, what I am getting at is this: Could you think of a
> > clever way of integrating content attributes for blobs
> > so that we could use it in NexusDB's SPs?
>
> We could have some nxtMimeBlob that saves attributes with
> the blob data either at the beginning or the end. In VCL
> there would be a function to get extra attributes, in SQL
> we would need a special UDF like e.g.
>
> select blobfield, BlobAttributes(blobfield) from table
>
> That sure would be possible

I agree that this might call for a new blob type; but,
then, there were times when blobs were unheard of.