i agree google groups is ugly

try it in fixed font mode and view as tree mode

it's  a little less ugly that way

Mark

Jerry Avins wrote:

> Richard Owlett wrote:
> 
>> robert bristow-johnson wrote:
>>
>>> testing...
>>>
>>> sheesh, this new Google Groups is weird....
>>>
>>> r b-j
>>>
>>
>> My ISP charges *more* than industry minimum.
>> For that I get:
>> 1. PORTAL-less WEB access
>> 2. Email with improving spam blocking and rock solid virus blocking
>> 3. access to newsgroups via supernews.com
>>
>> #3 is a biggee.
>> I don't see any of the problems many complain about.
>> I doubt that supernews.com is *ONLY* service that provides this 
>> advantage.
>>
>> *YOU DO GET WHAT YOU PAY FOR*  ;)
>>
>> PS. Haven't checked Google's privacy policy, But YAHOO's *STINKS*
> 
> 
> I'll do my own spam blocking, if you please. 

My ISP's system gives me best of both worlds. Virus blocking and spam 
*scoring* on their hardware. I get to chose what actually happens to a 
email with a specific score.

> Is the SwiftForth mailing list getting through to you?

It was always getting thru having a score of .47 on scale of 0-10 with 
my Quarantine level set at 1.0. Unfortunately some actual spam was 
scoring there or below. I'm lowering Quarantine level and tweaking 
whitelist.


> 
> AOL commercials imply -- hell: claim outright -- that you their service 
> is the only way to avoid catching a virus. I do my own virus checking, 
> too. Eben though IEEE and my ISP both scan for viruses, some have gotten 
> through to me and been squashed by Norton. A year of it costs about a 
> month of AOL. Meanwhile, I get news and web hosting for half of that 
> monthly.
> 
> Jerry

Richard Owlett wrote:
> robert bristow-johnson wrote:
> 
>> testing...
>>
>> sheesh, this new Google Groups is weird....
>>
>> r b-j
>>
> 
> My ISP charges *more* than industry minimum.
> For that I get:
> 1. PORTAL-less WEB access
> 2. Email with improving spam blocking and rock solid virus blocking
> 3. access to newsgroups via supernews.com
> 
> #3 is a biggee.
> I don't see any of the problems many complain about.
> I doubt that supernews.com is *ONLY* service that provides this advantage.
> 
> *YOU DO GET WHAT YOU PAY FOR*  ;)
> 
> PS. Haven't checked Google's privacy policy, But YAHOO's *STINKS*

I'll do my own spam blocking, if you please. Is the SwiftForth mailing 
list getting through to you?

AOL commercials imply -- hell: claim outright -- that you their service 
is the only way to avoid catching a virus. I do my own virus checking, 
too. Eben though IEEE and my ISP both scan for viruses, some have gotten 
through to me and been squashed by Norton. A year of it costs about a 
month of AOL. Meanwhile, I get news and web hosting for half of that 
monthly.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Richard Owlett wrote:
> robert bristow-johnson wrote:
>
> > testing...
> >
> > sheesh, this new Google Groups is weird....
> >
> > r b-j
> >
>
> My ISP charges *more* than industry minimum.
> For that I get:
> 1. PORTAL-less WEB access

dunno even what that is.  i dial in, i get access.

(BTW, i'm dialup because, even though i'm in the largest town in
Vermont, i live in the farthest northwest corner and am too far away
from the switch to get DSL and i hadn't wanted to get cable TV and
cable modem is not available without the cable TV hookup.  so dialup it
is.  don't send me any long attachements, please.)

> 2. Email with improving spam blocking and rock solid virus blocking

not much viruses out there for a Mac.

> 3. access to newsgroups via supernews.com
>
> #3 is a biggee.
> I don't see any of the problems many complain about.
> I doubt that supernews.com is *ONLY* service that provides this
advantage.
>
> *YOU DO GET WHAT YOU PAY FOR*  ;)

sometimes, i think that you get proportionately to what you pay for,
but that isn't even the case.  my ISP newserver doesn't seem too bad
(it's newsguy.com whoever those guys are), but some posts haven't shown
up.  NONE of these skewed OT posts have, and the original post hadn't,
but i could see it with google and news2web.com.

i used google groups (and its predicessor deja-news) in the past so i
could read and post to USENET remotely without dialing into my ISP from
home.  i've also used it to find old posts to read or to cite.  anyway,
this new beta version of Google Groups is not better, just weirder.
even though i was registered with the old, i had to re-register and
they asked for a nickname (and i'm glad i didn't say "shithead" or
similar because that's would have been posted in the From: header).
they didn't tell you.  (i fixed it, and am now posting this with
google.)  and they're doing all sorts of stupid unnecessary formatting
of the articles.  they're really taking a giant step backwards (in the
AOL direction) doing cutsie things when all we want is to read and post
unmolested.

> PS. Haven't checked Google's privacy policy, But YAHOO's *STINKS*

i don't mess with yahoo at all for anything.

r b-j

On Tue, 05 Apr 2005 17:20:23 -0700, Jon Harris wrote:
> On the topic of news service, I used to use and recommend news.individual.net,
> but just a week ago, they switched from being free to being a paid service.  Are
> there any good free news servers out there?  My ISP offers one so I'm back to
> using that.  It wasn't quite as good as news.individual.net, hence my switch,
> but it is decent.  News servers/readers are vastly superior to web access, IMHO.

I just ponied up to keep using news.individual.net.  It didn't seem like a
terribly large impost, and it does work pretty well.  My ISP news server
also works OK, but gets regularly black listed on account of spam input,
and is unavailable unless connected through said ISP (eg. at work...)

Running a news server is a pretty major undertaking.  I wouldn't expect it
to be done well for nothing.  Google gets its ad revenue, but you can't
force ads down an NNTP connection, so NNTP servers either have to be paid
for by your ISP contract or some other way.

Cheers,

-- 
Andrew

"Richard Owlett" <rowlett@atlascomm.net> wrote in message
news:1156a5c444l73bd@corp.supernews.com...
> robert bristow-johnson wrote:
>
> > testing...
> >
> > sheesh, this new Google Groups is weird....
> >
> > r b-j
> >
>
> My ISP charges *more* than industry minimum.
> For that I get:
> 1. PORTAL-less WEB access
> 2. Email with improving spam blocking and rock solid virus blocking
> 3. access to newsgroups via supernews.com
>
> #3 is a biggee.
> I don't see any of the problems many complain about.
> I doubt that supernews.com is *ONLY* service that provides this advantage.
>
> *YOU DO GET WHAT YOU PAY FOR*  ;)

On the topic of news service, I used to use and recommend news.individual.net,
but just a week ago, they switched from being free to being a paid service.  Are
there any good free news servers out there?  My ISP offers one so I'm back to
using that.  It wasn't quite as good as news.individual.net, hence my switch,
but it is decent.  News servers/readers are vastly superior to web access, IMHO.

robert bristow-johnson wrote:

> testing...
> 
> sheesh, this new Google Groups is weird....
> 
> r b-j
> 

My ISP charges *more* than industry minimum.
For that I get:
1. PORTAL-less WEB access
2. Email with improving spam blocking and rock solid virus blocking
3. access to newsgroups via supernews.com

#3 is a biggee.
I don't see any of the problems many complain about.
I doubt that supernews.com is *ONLY* service that provides this advantage.

*YOU DO GET WHAT YOU PAY FOR*  ;)

PS. Haven't checked Google's privacy policy, But YAHOO's *STINKS*

>OK,
>
>regarding fixe point scaling, here is some answer coming from a Texas
>Instrument Application Note that I have .
>I think they have some serious background in the domain of fixed point
>fft :-)
>
>  TI SPRA948 :
>Thus, the input data must be
>prescaled to allow for one bit of growth in the first two stages and
>two bits of growth in
>each stage thereafter if overflow is to be averted.
>
>Julien
>
>

Thanks, this is truly an interesting app note. If I read it carefully,
what it is saying is that in the first 2 stages you can get a growth of 2,
and all the next stages a growth of 2.414.
However I still do not belive that you need to compensate for 2.414 in
every last N-2 stages, simply because this 2.414 growth will not be
encountered in consecutive stages through 1 path.

Another way of reasoning:
Suppose a 256 point FFT. According to the paper this would have a growth
of 
2^2 * 2.414^6 = 792

However I think that a worst case input to my FFT would be when all output
would be on one FFT output point --> hence if my input is a pure sin/cos at
a certain frequency. For example a DC input (sin/cos of 0Hz). Obviously the
FFT output is just the sum of all the inputs. Hence I would conclude that
my maximum gain is 256 and not 792.


		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com

OK,

regarding fixe point scaling, here is some answer coming from a Texas
Instrument Application Note that I have .
I think they have some serious background in the domain of fixed point
fft :-)

  TI SPRA948 :
Thus, the input data must be
prescaled to allow for one bit of growth in the first two stages and
two bits of growth in
each stage thereafter if overflow is to be averted.

Julien

>Hi Newline,
>
> unless you have some knowledge on the input signal, you will have to
>protect against the worst case that could overflow your FFT; if your
>input signal is completely unknow, you may consider using a
>block-floating point FFT, i.e each stage adjust the scaling depending
>on the output of the previous stage.
>
>on the other hand, if you have some information about the input signal
>(for ex: OFDM modulator), just run a lot of simulation with the input
>signal to see the scaling you need at 1st stage then at 2nd stage,
>etc...
>
>Julien
>
>

Thanks for the response. 
I want to make a general purpose FFT, hence I know nothing of my input
signal. I agree I need to protect versus the worst case for overflow.

However my problem is exactly to find out what is the worst case.
The 1+sqrt(2) I have found a couple of times looks to me to be worser than
worst case. I think this because the 1+sqrt(2) comes from a pi/4 like
twiddle, but already in the first DIT FFT stage I have never a twiddle of
pi/4. So already at the output of my first stage I would have a bit too
much headroom. (and hence non optimal use of my bits)
Now if this is only valid in stage 1, probably there is not much to gain,
but I find it hard to figure this out for the next stages.






		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com