included files can get at them.

so we can do ESI, then gzip again to store it.

Add a vef_read() function which implements the esi_syntax&8 bit,
it is too late to do it when we call VEP_parse().

so it applies to all fetch methods.

via sp->wrk->storage.

to tuck in the last storage segment sitting on sp->work->storage,
rather than have this code duplicated througout all the VFP->end()'s

instance into sp->obj.

on reception

ESI+GZIP testcase

and return the offsets we need to compose the VEC string.

The default callback function is a 1:1 mapping.

so we can combine the CRC32 values.

object but wanting to store it ungzip'ed.

processing.

Add separate anchor fields for vgz for reception and esi work.
(The response vgz is a local variable)

Futher modelling of VGZ functions

Move the delivery gunzip away from session workspace and use a
64k stackbuffer for bufferspace instead.

Improve and clean up the VGZ allocation a bit.

Add a comment with our worst case gzip scenario

Now that the going is staring to get tough, split the ESI parsing from
the ESI fetch processing to keep source file size manageable.

Roll out the gzip/gunzip logic some more

Unify header handling for gzip, and make wrk->do_g[un]zip usable
as intention flags.

Minor cleanups

Retire the old ESI code.

Code polishing