sockets: htonl question

[Tom Impelluso]

Hello,

I am aware of the need for ntohl and htonl.
But floats?

I am aware of Beej's (Brian "Beej Jorgensen" Hall) implementation:

http://beej.us/guide/bgnet/output/html/singlepage/bgnet.html#htonsman

Yet, I devised my own kludge:
------------------------
uint32_t data, data1;
float num1;

i =read(descripto r, (char *) &data, sizeof(uint32_t ));
data1 = ntohl(data);
num1 = *(float*) ( &data1);
--------------------------

This works for me; always has!
Now, mind you, I have done scientific coding.
And in an enviroment with extremley well-behaved floats,
where round-off has never been an issue.
And very little exceptions....

So...

Why does my simpler implementation work?
What problems could it cause?
Friends have advised me to use Jorgensen, but I do not see the need.
(I also have not gotten it to work)

Thanks,
tom

[mlimber]

Re: sockets: htonl question

On Jul 23, 12:16 pm, Tom Impelluso <impel...@attil a.sdsu.eduwrote :

Hello,
>
I am aware of the need for ntohl and htonl.
But floats?
>
I am aware of Beej's (Brian "Beej Jorgensen" Hall) implementation:http://beej.us/guide/bgnet/output/ht....html#htonsman
>
Yet, I devised my own kludge:
------------------------
uint32_t data, data1;
float num1;
>
i =read(descripto r, (char *) &data, sizeof(uint32_t ));
data1 = ntohl(data);
num1 = *(float*) ( &data1);
--------------------------
>
This works for me; always has!
Now, mind you, I have done scientific coding.
And in an enviroment with extremley well-behaved floats,
where round-off has never been an issue.
And very little exceptions....
>
So...
>
Why does my simpler implementation work?
What problems could it cause?
Friends have advised me to use Jorgensen, but I do not see the need.
(I also have not gotten it to work)

Sockets are technically off-topic here (cf. <http://www.parashift.com/c
++-faq-lite/how-to-post.html#faq-5.9>).

<OT>I think "Beej" addresses your question as to why in section 6.4
<http://beej.us/guide/bgnet/output/html/singlepage/
bgnet.html#seri alizationwhen he mentions different binary formats
for floating point numbers.</OT>

Cheers! --M