Re: Uploading to Asimov

[pausch@saaf.se (Paul Schlyter)]

In article <ql4Ke.8225$p%3.35465@typhoon.sonic.net>,
Andy McFadden  <fadden@fadden.com> wrote:

> In comp.sys.apple2 Paul Schlyter <pausch@saaf.se> wrote:
>>> The real advantage of the compressed formats, IMHO, is less about
>>> compression and more about having a CRC on the files.  That takes
>>> the guesswork out of whether or not the images are damaged.
>> 
>> You don't need a compression program for that!  Instead you could
>> use an MD5 utility, which produces MD5 hashes from the file contents.
>> The MD5 hashes are then put in a separate ASCII file.  This method is
>> used on many other web sites.
> 
> I dislike this approach, for the following reasons:
> 
>  - Separate files can be lost.
>  - Separate files aren't always updated when the main file is updated.
>  - People may not check them.  If somebody downloads a file and it doesn't
>    work, they're more likely to post a "does this work" message than
>    go back and look for an MD5 file.  Having it checked automatically by
>    the unpacking program makes it impossible to avoid/ignore.

Separate files requires more of the user and the maintainer, that's true.
 
>> MD5 also provides better detection than CRC - not only because it has
>> more bits (MD5 has 128 bits, compared to 16 or 32 bits for the most
>> common CRC's), but because it detects not merely accidential
>> corruption but also deliberate modification of the files.
> 
> My concern in an Apple II archive isn't for deliberate corruption.
> 
> I agree that this is a general concern, and a cryptographically secure
> hash certainly beats a CRC.  If that's a real worry, though, MD5 gets
> you nothing by itself, because if I can replace the main file then I can
> also replace the MD5 file.  You need a digital signature.

A digital signature can be forged as easily as an MD5 hash.  To
really guard against deliberate tampering with the data, you need
more than that: you need a PKI, i.e. a digital certificate signed by
a CA you trust.

>> For ensuring the integrity of the files, MD5 is much superior to CRC.
> 
> Perhaps, but in this case it seems inconvenient and excessive.  If it
> were added to the Zip format that would be a good thing, but if all you're
> trying to detect is somebody using an ASCII transfer mode, a 16-bit CRC
> is probably good enough.

If you try to unpack a Zip file transferred by FTP in ASCII mode, you
probably won't even get as far as checking any CRC, because headers/etc
before that will get corrupt.

-- 
----------------------------------------------------------------
Paul Schlyter,  Grev Turegatan 40,  SE-114 38 Stockholm,  SWEDEN
e-mail:  pausch at stockholm dot bostream dot se
WWW:     http://stjarnhimlen.se/