[Info-vax] Re; Spiralog, RMS Journaling (was Re: FREESPADRIFT)

[Johnny Billquist]

On 2016-06-21 18:27, Jan-Erik Soderholm wrote:
> Den 2016-06-21 kl. 18:18, skrev VAXman- at SendSpamHere.ORG:
>> In article <nkbn1q$5cn$2 at Iltempo.Update.UU.SE>, Johnny Billquist
>> <bqt at softjar.se> writes:
>>> On 2016-06-19 04:35, David Froble wrote:
>>>> Jan-Erik Soderholm wrote:
>>>
>>>>> Yes, but that is a simple calculation with 1 block = 512 bytes.
>>>>> It does not correctly display partially filled blocks. I guess
>>>>> that what most here are talking about is the size up to the
>>>>> marker. Convering the used/allocated number of blocks is simple.
>>>>>
>>>>> Jan-Erik.
>>>>
>>>> I would suggest "what does it matter?"
>>>>
>>>> It's not like some other file is going to use the unused portion of
>>>> some
>>>> disk block, or, as suggested, unused part of 4096 bytes ....
>>>
>>> No. But if you implement FTP, or a web server, or any number of other
>>> tools/programs, they are expected to report file sizes in bytes, and the
>>> byte count should be *accurate*. Even one byte off is not acceptable.
>>
>
> A web server probably always has to calculate the number of bytes
> sent to get it right. There is many ways that the sent data can be
> different then the on-file data and the web page can be "dynamic".

It don't have to tell the size. But if you don't, you need to then 
either encode the transfer in a way that can indicate the transfer is 
complete, or close the connection at the end, to indicate that the 
transfer is complete. HTTO 1.0 didn't have any encoding trick, so there 
you had to close the connection, which is costly.

Also, without size information, you will not get any progress bars, or 
estimated time remaining. The browser will just show that it is working, 
and that is all you will know...

FTP is supposed to always be able to tell the correct size. The RFC also 
points out that on some systems, this can be very costly, so it 
shouldn't be requested lightheartedly. But Unix systems violate this 
both in calling it way more often, and also lying about the size.

>> Why?
>
> In HTTP you can end up with connections staying open and
> waiting for that final byte that never comes.

Yup. If you lie about the size, bad things will happen.

	Johnny