Ok, a few hours of trying various CPAN modules reveals the following working code:

use IO::Uncompress::AnyUncompress qw(anyuncompress)
my $input = `cat testfile.gz`;
my $output;
anyuncompress $input => \$output, { 'InputLength' => 1000 };
print $output;
[download]

I've confirmed that this works by benchmarking with vs. without the InputLength option ... using it yields about a 4x speedup on files of about 50KB (gzipped), since it's only uncompressing the leading 1000 characters.

Fyi, this is for a local web cache analysis program -- I need to periodically clean out certain files while leaving others. One of the conditions is whether the file is HTML or not, and you'd think you could rely on Content-Type headers, but as it turns out some web sites mis-type javascript and css content as text/html. Thus, I really need to inspect the file to be absolutely sure of the content type.

Anyways, problem is essentially solved at this point. Thanks to all for the help and suggestions!

cheers
MFN