This is part of The Pile, a partial archive of some open source mailing lists and newsgroups.
From: vantontl@my-deja.com
Subject: Re: Low cost web testing tools
Date: Mon, 28 Aug 2000 21:39:09 GMT
In article <39AA7757.820725E3@automatedtestsolutions.com>,
Donald Berrett <dberrett@automatedtestsolutions.com> wrote:
>
> We're looking for entry level web testing tools (functional
primarily).
>
> Anyone have any leads?
>
> Don Berrett
> --
>
>
If you're looking for entry level, try free. Here is a list of several
web-based web testing tool that will help you run some simple tests on
your site.
http://www.netmechanic.com/ (spelling, links, load time)
http://validator.w3.org/ (HTML validation)
http://www.cast.org/bobby/ (accessibility validation)
http://www.anybrowser.com/siteviewer.htm (browser compatibility)
http://ugweb.cs.ualberta.ca/~gerald/lynx-me.cgi (Lynx, text-only,
simulator)
===
From: 11cents <11cents@my-deja.com>
Newsgroups: comp.software.testing
Subject: Re: Low cost web testing tools
Date: Mon, 28 Aug 2000 21:52:22 GMT
In article <39AA7757.820725E3@automatedtestsolutions.com>,
Donald Berrett <dberrett@automatedtestsolutions.com> wrote:
>
> We're looking for entry level web testing tools (functional
primarily).
>
> Anyone have any leads?
"Low cost" is a relative term. How low do you mean? Also, some software
is licensed per seat, so how many people do you need to run the
software? What level of testing do you want. Many tools may have less or
more functionality than you need.
Mercury's Astra is a "lower" cost tool (lower than WinRunner, that is).
RSW's e-Tester (in my opinion) is better because it hooks right into
their e-Load tool that kicks Astra's ass.
There's a new-comer on the block called eValid, but I think it is
marketed more as a load tool than a functional/regression testing tool.
===
From: Ralf Blazek <rblazek@bigfoot.de>
Newsgroups: comp.software.testing
Subject: Re: Web Testing with Rational
Date: Mon, 28 Aug 2000 22:05:57 +0200
11cents wrote:
>
> I think in general, Rational makes decent software, but they don't
> bubble to the top of my list when I think of tools for e-business web
> testing. The top of that list would be the likes of RSW and Mercury. Why
> not shop around? Evaluate some of the tools out there first-hand and see
> if they fit your need. Since you are a small group, probably RSW's
> e-Test Suite will fit better into your organization than
> WinRunner/LoadRunner. But, don't take my word for it, most tools out
> there will allow evaluation periods. Use this to decide for yourself.
Thank you for your comments. I have evaluated the tools you mentioned
and I would agree with you, that RSW e-Test is an outstanding tool for
functional web testing. But they have no bug tracking tool! I have to
communicate all the defect to the developers. Presently I do this with
small WORD documents send by e-mail, which is a great pain. As much as I
have seen Rational ClearQuest would be a great help.
Does anybody use RSW e-Test with some bug tracking tool? How good do
they fit together? Or is Rational TeamTest the better solution?
===
From: egyptian_magician@my-deja.com
Newsgroups: comp.software.testing
Subject: OK to use Single computer for Web Load Testing ?
Date: Wed, 30 Aug 2000 00:00:34 GMT
Greetings,
Is there any disadvantage to using a single computer with high ram to
perform web load testing , rather than having seperate client machines?
Also, I am of the impression that as far as web surfing is concerned,
the processor speed is not signifiant. is this correct?
===
From: "Dan" <none@nospam123.com>
Newsgroups: comp.software.testing
Subject: Re: OK to use Single computer for Web Load Testing ?
Date: Tue, 29 Aug 2000 19:58:01 -0700
<egyptian_magician@my-deja.com> wrote:
> Is there any disadvantage to using a single computer with high ram to
> perform web load testing , rather than having seperate client machines?
>
> Also, I am of the impression that as far as web surfing is concerned,
> the processor speed is not signifiant. is this correct?
Not to be evasive, but I think your question is too general
to get any sort of useful answer.
You'll need to describe what you wish to load test... Do you
have lots of web servers (a web farm)? What are they? How
many hits per day/hour/minute/second (take your pick) would
you like the servers to be able to handle? Do your web
servers provide dynamic content, or just static pages? If
static pages, are there lots of big graphics and file
downloads? What is your pipe to the net, T3, T1?
There are several disadvantages to using a single machine to
generate load:
1_ It is not a realistic representation of real world HTTP loads. Lots of
machines will hit your server(s), not just one.
2_ Ram is only one small part of the equation in generating load. If you
have one machine, you only have one network card. One network card can't
possibly generate enough load, unless your server(s) are really really slow.
If your site is going to be using large graphics and downloads, or maybe
dynamic data, you one load client is going to be network, processor, disk
and ram bottlenecked..
Is the one machine your describing a 8 processor Unix system with a gig of
memory and multiple ethernet cards, or a Pentium II with a 366 and 128megs
of ram? What kind of load testing software are you using?
===
From: egyptian_magician@my-deja.com
Newsgroups: comp.software.testing
Subject: Re: OK to use Single computer for Web Load Testing ?
Date: Wed, 30 Aug 2000 06:16:20 GMT
In article <eO_q5.1048$mQ2.119248@news.pacbell.net>,
"Dan" <none@nospam123.com> wrote:
> Hi!
> Not to be evasive, but I think your question is too general to get any
> sort of useful answer.
>
> You'll need to describe what you wish to load test...
Thanks so much for the response. I'm a newbie so bear with me.
I wish to test an e-mail web application in the staging environment.
We get about 20,000 users per hour. Average user session is about 10
minutes . So we have about 20K / 6 = 3333 users logged
in at a given 10 minute interval.
The network engineers told me there is a 10:1 scalability factor
between production and staging environment. ( I believe there are
10 production web servers and 1 staging web server. )
So I figure 333 virtual users on staging is an adequate simulation .
> 1_ It is not a realistic representation of real world HTTP loads. Lots
> of machines will hit your server(s), not just one.
Sorry. I am lost on this point.The material I have read so far from
from Mercury, RSW ,E-Valid indicate that a single machine can perform
the load test.
Here is a newsgroup quote on the topic from the E-valid vendor:
"
Each eValid browser has a ~581KB footprint. One can run many hundreds
of copies of eValid on a fast PC. We have run over 250 browsers on a
256MB, 550 MHz Pentium III running Windows 2000. CPU utilization is
not that high, it turns out. However, we did have to modify the desktop
heap setting and the amount of virtual memory available.
These facts are discussed in some detail at this URL:
<http://www.soft.com/eValid/Products/Documentation.2/load.unlinked.html>
"
So I was thinking a single pentium 3 proc machine, 512k ram, single
network card will work.
>One network card can't possibly generate enough load, unless your
>server(s) are really really slow.
I am a hardware novice . I don't follow why one network card
can't do this.
> Do your web servers provide dynamic content, or just
> static pages? If static pages, are there lots of big graphics and >
>file downloads? What is your pipe to the net, T3, T1?
We have a T3 connection. Static pages with very little graphics and
light file download.
> There are several disadvantages to using a single machine to generate
>load:
>
> 1_ It is not a realistic representation of real world HTTP loads. Lots
> of machines will hit your server(s), not just one.
I think I remember hearing that these load tools will simulate different
ip addreses. Even if they didn't , I'm still not sure why it matters.
A connection is a connection isn't it? - regardless whether it is
coming from a single machine or multiple machines.
Also, what hardware would you suggest to do my load test?
Multiple network cards? etc
I will be using a Windows NT machine, since this is the most common
platform that our customers use.
Our servers are Sun Solaris running Apache Web Server. Oracle DB.
===
From: egyptian_magician@my-deja.com
Newsgroups: comp.software.testing
Subject: Re: OK to use Single computer for Web Load Testing ?
Date: Wed, 30 Aug 2000 06:16:20 GMT
In article <eO_q5.1048$mQ2.119248@news.pacbell.net>,
"Dan" <none@nospam123.com> wrote:
> Hi!
> Not to be evasive, but I think your question is too general to get any
> sort of useful answer.
>
> You'll need to describe what you wish to load test...
Thanks so much for the response. I'm a newbie so bear with me.
I wish to test an e-mail web application in the staging environment.
We get about 20,000 users per hour. Average user session is about 10
minutes . So we have about 20K / 6 = 3333 users logged
in at a given 10 minute interval.
The network engineers told me there is a 10:1 scalability factor
between production and staging environment. ( I believe there are
10 production web servers and 1 staging web server. )
So I figure 333 virtual users on staging is an adequate simulation .
> 1_ It is not a realistic representation of real world HTTP loads. Lots
> of machines will hit your server(s), not just one.
Sorry. I am lost on this point.The material I have read so far from
from Mercury, RSW ,E-Valid indicate that a single machine can perform
the load test.
Here is a newsgroup quote on the topic from the E-valid vendor:
"
Each eValid browser has a ~581KB footprint. One can run many hundreds
of copies of eValid on a fast PC. We have run over 250 browsers on a
256MB, 550 MHz Pentium III running Windows 2000. CPU utilization is
not that high, it turns out. However, we did have to modify the desktop
heap setting and the amount of virtual memory available.
These facts are discussed in some detail at this URL:
<http://www.soft.com/eValid/Products/Documentation.2/load.unlinked.html>
"
So I was thinking a single pentium 3 proc machine, 512k ram, single
network card will work.
>One network card can't possibly generate enough load, unless your
>server(s) are really really slow.
I am a hardware novice . I don't follow why one network card
can't do this.
> Do your web servers provide dynamic content, or just
> static pages? If static pages, are there lots of big graphics and >
>file downloads? What is your pipe to the net, T3, T1?
We have a T3 connection. Static pages with very little graphics and
light file download.
> There are several disadvantages to using a single machine to generate
>load:
>
> 1_ It is not a realistic representation of real world HTTP loads. Lots
> of machines will hit your server(s), not just one.
I think I remember hearing that these load tools will simulate different
ip addreses. Even if they didn't , I'm still not sure why it matters.
A connection is a connection isn't it? - regardless whether it is
coming from a single machine or multiple machines.
Also, what hardware would you suggest to do my load test?
Multiple network cards? etc
I will be using a Windows NT machine, since this is the most common
platform that our customers use.
Our servers are Sun Solaris running Apache Web Server. Oracle DB.
===
From: opc73@my-deja.com
Newsgroups: comp.software.testing
Subject: Re: OK to use Single computer for Web Load Testing ?
Date: Wed, 30 Aug 2000 19:42:36 GMT
In article <8oi8r8$cni$1@nnrp1.deja.com>,
egyptian_magician@my-deja.com wrote:
> >One network card can't possibly generate enough load, unless your
> >server(s) are really really slow.
>
> I am a hardware novice . I don't follow why one network card
> can't do this.
It's a function of hardware/bandwidth limitation. you can
only make so many network service requests before the NIC
card becomes the bottleneck in your simulated system. this
bottleneck will limit the amount of traffic to the site and
typically, this amount of traffic will be less than the
actual load you want to test. if it's not, you should be
concentrating on developing a marketing strategy first
before you spend resources on load testing ;-)
> > 1_ It is not a realistic representation of real world HTTP loads.
Lots
> > of machines will hit your server(s), not just one.
>
> I think I remember hearing that these load tools will simulate
different
> ip addreses. Even if they didn't , I'm still not sure why it matters.
>
> A connection is a connection isn't it? - regardless whether it is
> coming from a single machine or multiple machines.
no. typically a complex web application will differentiate
between machines accessing it's resources so that they can
track individual user data via IP addresses. this can be
accomplished in a relatively simple fashion with cookies, or
some sort of middle tier software layer like com objects/SQL
backend for more sophisticated purposes. therefor you're
exercising different code when you connect with different IP
addresses(machines) and putting significantly more load on
the application.
===
From: "Dan" <none@nospam123.com>
Newsgroups: comp.software.testing
Subject: Re: OK to use Single computer for Web Load Testing ?
Date: Wed, 30 Aug 2000 21:13:14 -0700
<egyptian_magician@my-deja.com> wrote:
> In article <eO_q5.1048$mQ2.119248@news.pacbell.net>,
> "Dan" <none@nospam123.com> wrote:
> > Hi!
> > Not to be evasive, but I think your question is too general to get any
> > sort of useful answer.
> >
> > You'll need to describe what you wish to load test...
>
> Thanks so much for the response. I'm a newbie so bear with me.
> I wish to test an e-mail web application in the staging environment.
>
>
> We get about 20,000 users per hour. Average user session is about 10
> minutes . So we have about 20K / 6 = 3333 users logged
> in at a given 10 minute interval.
>
> The network engineers told me there is a 10:1 scalability factor
> between production and staging environment. ( I believe there are
> 10 production web servers and 1 staging web server. )
>
> So I figure 333 virtual users on staging is an adequate simulation .
>
> > 1_ It is not a realistic representation of real world HTTP loads. Lots
> > of machines will hit your server(s), not just one.
>
> Sorry. I am lost on this point.The material I have read so far from
> from Mercury, RSW ,E-Valid indicate that a single machine can perform
> the load test.
>
>
> Here is a newsgroup quote on the topic from the E-valid vendor:
> "
> Each eValid browser has a ~581KB footprint. One can run many hundreds
> of copies of eValid on a fast PC. We have run over 250 browsers on a
> 256MB, 550 MHz Pentium III running Windows 2000. CPU utilization is
> not that high, it turns out. However, we did have to modify the desktop
> heap setting and the amount of virtual memory available.
>
> These facts are discussed in some detail at this URL:
> <http://www.soft.com/eValid/Products/Documentation.2/load.unlinked.html>
> "
>
> So I was thinking a single pentium 3 proc machine, 512k ram, single
> network card will work.
>
>
> >One network card can't possibly generate enough load, unless your
> >server(s) are really really slow.
>
> I am a hardware novice . I don't follow why one network card
> can't do this.
>
> > Do your web servers provide dynamic content, or just
> > static pages? If static pages, are there lots of big graphics and >
> >file downloads? What is your pipe to the net, T3, T1?
>
> We have a T3 connection. Static pages with very little graphics and
> light file download.
>
> > There are several disadvantages to using a single machine to generate
> >load:
> >
> > 1_ It is not a realistic representation of real world HTTP loads. Lots
> > of machines will hit your server(s), not just one.
>
> I think I remember hearing that these load tools will simulate different
> ip addreses. Even if they didn't , I'm still not sure why it matters.
>
> A connection is a connection isn't it? - regardless whether it is
> coming from a single machine or multiple machines.
>
>
> Also, what hardware would you suggest to do my load test?
> Multiple network cards? etc
>
> I will be using a Windows NT machine, since this is the most common
> platform that our customers use.
>
> Our servers are Sun Solaris running Apache Web Server. Oracle DB.
Great! This gives me a little more to chew into :-)
Even though E-Load (and others) claim that you can run 250 simulated users
on a single PC, the more important thing is knowing exactly what each of
those users are *doing*... All load test tools have one thing in common:
they start multiple threads or processes, and each thread or process acts as
a single simulated user. So, while it may very well be possible to *start*
250-300 (or more) threads or processes on a single machine, as soon as each
of those simulated users starts to do some logic, such as downloading a JPEG
file, verifying content of a web page, checking dynamic data, you can bet
that the speed with which those users run decreases, and the system
resources taken by them increases . I further suspect that the reason E-Load
(and others) claim that it's processor was not pegged during testing is
because the 250 virtual users weren't doing very much! Add some dynamic data
and some file downloads, and a Pentium III will be pegged trying to generate
much less load.
It's like when you read the product literature published by an automobile
manufacturer about a car they produce, they typically mention a "top speed".
What they don't mention is that they had a professional driver, a competely
flat racetrack, super high octane fuel, and a tail wind to achieve top
speed. Simply put, the claims of Mercury, E-load, and others (while not at
all deceptive) are simply not applicable, since they usually take the best
case scenerio. At best they can be used for a very rough estimate.
With that said, the bad news is that no one can tell you with any fine level
of accuracy whether the load you wish to achieve can be done with a single
machine. They can guess, but the fact is that you won't know until you try.
The reason no one will tell you, is because of the HUGE number of variables
present. The OS you are running, how that OS is tuned, what kind of ethernet
card you have, what kind of traffic you have on your network, how fast your
web servers are, how the web server software is tuned, what kind of data
your web clients are supposed to generate... The list goes on and on. The
only reason I guessed is because one machine is almost never sufficient but
for the smallest of sites... Yours does not appear to be small... Read on..
So, how are you supposed to figure out how much horsepower you need? I
recommend that you start testing with a single machine, and see how much
load it can generate before one of it's components (memory, OS, network)
gets overloaded. Then add resources (more machine etc..) as needed... My
best guess based on the numbers you gave (333 virtual users generating 20k
each) is that when you start your load test with 333 virtual clients, your
load client will be network I/O bound. In other words, the network interface
will be under duress on the single load client, and will not be able to take
advantage of the 333 virtual users that are trying to generate load. I
think the suitespot of number of simulated users to available system
resources will be much lower. I also think that with that number of
simulated users running, your operating system will be under great duress
just coping with juggling 333 seperate threads/processes... If each of those
simulated clients is asked to do any sort of complicated logic, then you can
watch your processor and memory soar...
I tell my customers to start with a single (preferably a loaner) machine...
Then, design a load scenario by recording a typical session. Then, run a
CPU, Process, memory, and network monitor program on both the server(s) and
the clients. If you have an ethernet sniffer, it would be great to employ
that as well, but that's expensive. Start with a small amount of simulated
users, and increase it to the point where one element (network, memory,
etc...) is seen to be a bottleneck. You can then decide to scale up the
number of machines, amount of memory based on what you see. It's a dynamic
process, this tuning.
Also, realize that your scaling factor may not give you an accurate view of
what will happen in production. In fact, I doubt it will. If you have 3,333
user logged in and using your system, and you intend to test using a 'test
environment' which simulates a small fraction of that (in this case, 10%),
or 333 users, there is no guarentee that your system will not experience
unique bottlenecks in when 3,333 users hit your site realtime... In other
words, a 10:1 scaling ratio leaves too much margin for error. As you scale
up, lots of other things come in to play. For instance, if your server can
handle 333 simulated users from one load client with one ethernet interface,
but that ethernet interface is only able to deliver a fraction of the load
that the simulated users are generating (say, 100 of those simulated users
actually can generate load), then you are going to get hammered when 3,333
unique computers (real users) use your site, because you've only really
tested the effect of 100 users...
Also, we haven't breached the topic of concurrent users versus nonconcurrent
users. I wrote a news posting a couple of months ago that explains this..
It's also important... Check www.dejanews.com and search for 'dpeknik' to
find it...
===
From: evalid <efmiller@pacbell.net>
Newsgroups: comp.software.testing
Subject: Re: OK to use Single computer for Web Load Testing ?
Date: Thu, 31 Aug 2000 15:44:12 -0700
egyptian_magician@my-deja.com wrote:
> Is there any disadvantage to using a single computer with high ram to
> perform web load testing , rather than having seperate client machines?
>
> Also, I am of the impression that as far as web surfing is concerned,
> the processor speed is not signifiant. is this correct?
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
So far as we can determine, given that you want to do realistic full-
browser tests that really do download every base page and all of its
components, there is no difference in terms of successfully imposing
load if you use multiple browsers all running on a "high ram" machine.
But is is fair to point out that if you have two or more browsers
running simultaneously and automatically retrieving pages, you probably
have to make sure that the cache function is suppressed. This would be
true especially if you were running the same script many times. But,
given that you want total base-page and component download times you
probably want the cache function suppressed in any case.
We hope this is a useful contribution to the discussion. Complete
details on the eValid solution can be had from http://www.e-valid.com
and free keys for EVALs -- which normally include the LoadTest
capability -- are available at <license@soft.com>.
Note: This response was posted by the eValid support team.
===
From: Donald Berrett <dberrett@automatedtestsolutions.com>
Newsgroups: comp.software.testing
Subject: How is Java code tested (in a Web page)?
Date: Wed, 30 Aug 2000 12:52:21 GMT
This is probably a DQ, but how do the automated test systems actually
test the Java in a web page?
===
Date: Fri, 1 Sep 2000 07:56:59 -0500
From: Vemoc <vemoc@hotmail.com.nospam>
Subject: [General Discussion] (Astra Quick Test) Dynamic Link on web page
Newsgroups: comp.software.testing
This is my code:
Sub SearchString()
Dim Texte
Dim Doc
Browser("ClickonJob.com").Sync
Set Doc = Browser "ClickonJob.com").Page("ClickonJob.com")
Texte = Doc.Link(1).QueryValue("text")
MsgBox texte
Doc.Link(Texte).click
End Sub
Call SearchString
My question now !!!
1- Why I can't put 1 to select the first link
2- Why I can't use a variable for select this link.
What I want to do, is select the first, second or other link in my web page. But the problem is: the text for the links are changing because they are dynamics. So is it possible to select the first link no matter what this link is ???
I'm open to any suggestions...
Thanks
===
From: evalid <efmiller@pacbell.net>
Newsgroups: comp.software.testing
Subject: Re: [General Discussion] (Astra Quick Test) Dynamic Link on web page
Date: Fri, 01 Sep 2000 13:03:28 -0700
Vemoc wrote:
> ...snip...
>
> What I want to do, is select the first, second or other link in my web page. But the problem is: the text for the links are changing because they are dynamics. So is it possible to select the first link no matter what this link is ???
>
> I'm open to any suggestions...
> Thanks
When the text for a link is changing dynamically,
or the location of the link on the page is changing
dynamically, you can exploit a feature ineValid called
"adaptive playback" to make your playback scripts very reliable.
This capability overcomes this very common kind of problem
by adapting what it does to fit the request...even in case the
text on the link has changed [but not if the linked URL has
changed], or in case the sequence of the link on the page
has changed [but notif the link is not on the page at all].
Try eValid's DEMO from http://www.e-valid.com or get a
full-function evaluation key from <licenses@soft.com>.
Note: This response posted by the eValid support team.
===
From: egyptian_magician@my-deja.com
Newsgroups: comp.software.testing
Subject: Opinions on Radview Webload?
Date: Tue, 05 Sep 2000 12:06:00 GMT
Anyone have experience/comments on the Webload tool?
I know it has less features than other tools and is only http recorder
But it is relatively inexpensive compareed to RSW and Mercury.
===
Date: Tue, 5 Sep 2000 19:55:14 -0500
From: knaucqos <knau@cqos.com.nospam>
Subject: [SilkTest] Difficulities recognizing web popup windows
Organization: mail2news@nym.alias.net
Newsgroups: comp.software.testing
I am currently using silktest 5.01 to test my company's web
application on IE 5.0. Currently i have a problem with
Silktest not properly recognizing a popup window during
playback. This popup window is used to display some graphs
and tables. The windows identifier sees the popup window as:
DialogBox("Configuration Report - Microsoft Internet
Explorer|$explorer5").BrowserChild("#1)
Now if I use this in my code for example like this:
DialogBox("Configuration Report - Microsoft Internet
Explorer|$explorer5").BrowserChild("#1).SetActive () //this
is also what the recorder would have written down
an then just run a simple test case that includes that line an error message will appear:
[ ] *** Error: Window '[DialogBox]$explorer5' was not found
[ ] Occurred in SetActive
This is obviously very strange. Silk recognizes the popup
window with the windows identifier and the test recorder,
but when playing back the testscript it fails to see the
popup window.
Please click the following link to reply:
http://www.qaforums.com/boards/ubb/Forum1/HTML/001003.html
===
From: "DJGray" <DJGray@attachmate.com>
Newsgroups: comp.software.testing
Subject: Re: [SilkTest] Difficulities recognizing web popup windows
Date: Wed, 6 Sep 2000 06:56:19 -0700
"knaucqos" <knau@cqos.com.nospam> wrote:
> I am currently using silktest 5.01 to test my company's web application on
> IE 5.0. Currently i have a problem with Silktest not properly recognizing a
> popup window during playback. This popup window is used to display some
> graphs and tables. The windows identifier sees the popup window as:
>
> DialogBox("Configuration Report - Microsoft Internet
> Explorer|$explorer5").BrowserChild("#1)
>
> Now if I use this in my code for example like this:
> DialogBox("Configuration Report - Microsoft Internet
> Explorer|$explorer5").BrowserChild("#1).SetActive () file://this is also
> what the recorder would have written down
>
> an then just run a simple test case that includes that line an error
> message will appear:
> [ ] *** Error: Window '[DialogBox]$explorer5' was not found
> [ ] Occurred in SetActive
>
> This is obviously very strange. Silk recognizes the popup window with the
> windows identifier and the test recorder, but when playing back the
> testscript it fails to see the popup window.
>
> Please click the following link to reply:
> http://www.qaforums.com/boards/ubb/Forum1/HTML/001003.html
knaucqos,
This sounds real similar to a problem I encountered several years ago with
SILK. (Haven't tested with a browser for a while, so this is a vague memory)
If I recall correctly, there were two browser windows open and I needed to
switch between them. I seem to remember that SILK was recognizing the
browsers differently depending on which one was active. For example, we have
browser A and browser B. If B is active, then it was considered #1 by SILK
and A was #2, so I did a set active on #2 (A) and now it became #1 and Bwas
#2.
Like I said, this was years ago, but I remember how frustrating it was till
I figured out what SILK was doing. This procedure may not be entirely
accurate, but it might give you enough to track down the issue. Good luck!
===
Date: Tue, 5 Sep 2000 20:06:14 -0500
From: knaucqos <knau@cqos.com.nospam>
Subject: [SilkTest] logging into your web application.
Newsgroups: comp.software.testing
Using netscape or Ie5.5, silk test logs into our web app
with no problem, but in IE5.0 it is unable to log in. Is
there a way to get it to log in?
Please click the following link to reply:
http://www.qaforums.com/boards/ubb/Forum1/HTML/001006.html
===
Date: Wed, 13 Sep 2000 04:32:21 -0500
From: patrick <pconlon@closerlook.com.nospam>
Subject: [General Discussion] Q: Build Process in a web environment
Newsgroups: comp.software.testing
An open question to those of you who work in software
quality assurance in a web environment. Does QA have
responsibility for documenting the build process or does
development? Does QA actually do the periodic builds or
does development? Whichever it is, do you agree with how
your company is doing things or would you prefer to do
things differently?
Please click the following link to reply:
http://www.qaforums.com/boards/ubb/Forum15/HTML/000249.html
===
From: "Oren Benkin" <oren_b@attglobal.net>
Newsgroups: comp.software.testing
Subject: Re: [General Discussion] Q: Build Process in a web environment
Date: Wed, 13 Sep 2000 22:48:45 +0200
All documentation and periodic builds responsibility is
R&D's. The QA is solely quality assurance, not development.
===
From: egyptian_magician@my-deja.com
Newsgroups: comp.software.testing
Subject: Web Server Response time & Load testing
Date: Mon, 18 Sep 2000 22:32:55 GMT
As I understand it, web server response time can be defined as
the time from the client browser making a request to when the
first byte of data starts to transfer.
Suppose I am load testing a server which can handle thousands
of concurrent users.
Also, I have a NT load generating machine capable of running 1- 500
clients all running the same sequence of actions.
What happens to the response time (as defined above) as the # of
load clients increase?
I'm not sure how the NT operating system works.
If client #1 makes a page request and is waiting for a server response,
does client # 2 have to wait before its request gets executed
or does the OS switch out client #1 process and execute the client #2
request?
===