L-Soft seeks partner to provide "off the shelf" mailing list service
--------------------------------------------------------------------
In order to provide the many Internet users who cannot afford expensive
IBM hardware with cost-effective mailing list solutions, L-Soft
international plans to start an "off the shelf" mailing list service.
Various levels of support would be provided to better meet the broad
needs of the LISTSERV community. Organizations with existing LISTSERV
expertise that plan to phase out their IBM hardware in the short term
will be able to retain the level of service their users were used to by
simply contracting L-Soft to run the lists on a central L-Soft system.
Such customers would retain full control over the administration of the
list and enjoy the lowest service fees, because L-Soft's manpower costs
would be minimal. On the other hand, an organization without any LISTSERV
(or computer) expertise, such as a Human Sciences department, could
contract L-Soft to take full responsibility for the operation and
administration of the list, in essence providing a "LISTSERV that you
program in English" service which would probably be cheaper than a unix
workstation on which to run a mailing list manually.
Since LISTSERV will not be available on cheap hardware within the next
three months, this service will have to be started on a VM system.
Purchasing a dedicated IBM mainframe would not be cost effective and
would lead to a steep price schedule, in effect putting the service out
of reach of most academic institutions. The investment would be in the
hundreds of thousands, whereas only a few percent of the CPU would be in
use. The P/370 is not currently available in a suitable configuration
(with TCP/IP) and may not scale up to the expected I/O load (the limiting
factor is the CPU power of the PS/2 host, which cannot be increased at
will).
Running the service on existing hardware with spare capacity, on the
other hand, could be cost effective while at the same time improving
quality of service. A large system with spare cycles would provide a
better response time on the average than an entry-level dedicated system,
especially where I/O is concerned. In addition, an existing system with
thousands of users will have much better operational facilities (backups,
operator coverage, etc) than L-Soft could afford for a dedicated system.
Finally, a typical side effect of downsizing plans is that spare cycles
become available as applications are moved away little by little, whereas
the operational costs remain more or less constant. The hardware cannot
be removed until the last administrative applications have been
relocated, and meanwhile there is no use for the spare cycles within the
organization. They can be rented easily as the users will not complain
about the extra load, but finding a customer willing to move his work to
a machine that will be removed within 1-2 years may not be easy, so the
charges are likely to be reasonably low.
L-Soft would like to invite interested parties to submit bids for the
provision of the service described below. While we understand that it may
take time to contact all the people who need to be involved in the
preparation of a formal bid, we would like a first contact to be
established before December 20th, even if this is on a non-binding basis
and some information is missing. Some sites may be in a position to make
a binding offer before January and our customer feedback suggests that
this service is needed urgently, so we will need to know what are the
other outstanding proposals in order to decide whether to wait or take up
one of the offers already made. L-Soft is in a position to start this
service as early as mid January, although it would be unrealistic to
expect all the paperwork to be cleared on such short notice.
*************************
* Hardware requirements *
*************************
Hardware requirements are given in terms of IBM model numbers, but third
party equipment is equally acceptable.
Ideally, the CPU should be a 9121, 9021 or ESA-capable 3090. We will also
consider bids based on a 9221-170 (or higher) or ESA-capable 4381. We
expect to need at most 12 hours of 3090-equivalent time a month for the
first year.
Any DASD model with a transfer rate of 3M/sec or higher is acceptable. As
long as the unit and path are not saturated, performance is not really an
issue and we would rather get the cheapest unit type (eg 3380 rather than
cached 3390). We expect to need a couple hundred megabytes to a gigabyte
of disk space for the first year, depending on customer demand. We are
willing to consider arrangements in which we would purchase a secondhand
3380 unit that you would connect to your system. L-Soft would then use 2
volumes and you would be allowed to use the other 2 as you see fit.
Unless we can get a long-term commitment, we will be running a second
level VM system so that we can easily move the service to another machine
at the end of the contract period. This means we cannot use SFS space
unfortunately.
*************************
* Software requirements *
*************************
The machine must run VM/ESA 1.1 or higher, TCP/IP V2.2 and RSCS V3. While
not absolutely mandatory, VS PASCAL is highly desirable.
******************************
* Operational considerations *
******************************
The system must be available 365 days a year, except for scheduled
relocation and the like. Ideally we would like 24x7 operator coverage,
although we will settle for any reasonable arrangement that somehow
addresses unexpected problems during the weekend. This point should be
explicitly covered in the bid. We will need to guarantee a certain level
of availability to corporate customers and we must have reasonable
assurance that the system cannot remain down for 4 days during extended
weekends, as we would lose money.
The system should have an uptime of 99.3% or better on a monthly basis,
ie at most 5 hours of downtime, scheduled or not, not counting hurricanes
and the like. We will not want a formal guarantee as long as we can
terminate the agreement if the system consistently fails to meet this
requirement. We will accept longer downtime on an exceptional basis to
relocate equipment, install upgrades, and so on.
You will be responsible for backing up our data at least once a week. You
will restore our data at no charge in case of hardware problem or any
other problem beyond our control. You will restore data at our expense if
we delete a file by mistake and need to have it reloaded. You will not be
liable for consequential damage if you lose our data, but if you lost
more than one day's worth we will not pay for the usage of the disk space
since the last working backup or for the last month, whichever is more.
We will also want to be able to exit the contract if you lose more than
one week's worth.
Because we will be using a second-level system you need not worry about
backing up our first-level spool files. Backup of our second-level spool
files will be our responsibility. You will provide a SPTAPE drive upon
request and at no extra charge at times when we must make system
generations: twice a year for time zone changes if you are running VM/ESA
1.1, whenever you relocate the hardware, and whenever you require us to
upgrade to a newer version or service level of VM.
************************************
* System management considerations *
************************************
We will manage the second-level system and your only responsibility will
be to configure the first-level userid properly and provide us with all
the information we need to set up the second-level system. Since we
expect to take a R/O link to public disks such as MAINT 190 (which you
are presumably caching), we will need to be informed of your procedures
and schedules for updating such public disks, so that we can take the
necessary steps to protect our second-level applications.
We will work together to minimize the impact of our second-level workload
on your first-level applications, and optimize the response time of the
second-level system. While we do not require V=F regions or the like, the
bid should describe the kind of configuration you are offering. In
particular we want to know how much first-level storage the second level
system will have, whether we will have a SET RESERVE or locked frames to
avoid dual paging, what SHARE we will get, and so on. We expect to run
the second-level system in a 10-16M virtual machine, depending on how
storage-constrained your first-level system is. At 16M there should be
virtually no paging.
****************
* Connectivity *
****************
The system must have excellent Internet connectivity with reasonable RTT
and packet loss rates (ie IP over X.25 will be a strong deterrent). We do
not expect to use much bandwidth as most of our traffic will be mail. For
pseudo-legal reasons we will not accept to be charged for connectivity by
volume, as we want to avoid anything that might give the impression that
we are selling connectivity or network access.
The second-level system will need its own IP address and hostname (again
so that it can be easily relocated when the contract runs out). This can
be accomplished either via a dedicated 8232/BTI, if you have a spare one,
or via a connection to the first-level TCPIP. We will take care of
registering the host and providing primary name service, but we may
require you to provide secondary name service on a local machine for
better reliability.
The second-level system will have to be connected to BITNET. If your
first-level system is a BITNET node with VMNET (or other high-speed
connection) we will simply use a VCTC connection to the first-level RSCS
as that is the most convenient and cost-effective solution. In that case
it will be necessary (for administrative reasons beyond our control) for
you to take responsibility for managing our BITNET node registration with
the NCC or BITNIC, as appropriate. We will pay for any increase in
membership charges caused by our connection. EARN sites should note that
we are authorized to connect L-Soft systems to EARN at no charge until
July 1996, and such connections do not count towards volume-based country
membership charges. In other words, there is no extra charge for EARN
sites, neither to the institution hosting the connection nor to its
national network. If the first level system is not connected to BITNET,
we will license VMNET at our expense and establish a link to a suitable
system. In that case we will take care of all aspects of the connection.
***********
* Charges *
***********
The agreement should be for an initial period of one year, preferrably
with an option to renew it on a quarterly basis after that. We will need
to know how long you expect to be able to offer the service.
We must be explicitly authorized to resell the resources you are selling
us. While we will probably not sell second-level accounts, we will charge
our customers for the disk space allocated to list archives and the like,
and we must have written permission to do such things.
We will need some form of protection against accidental CPU charges. For
instance, if the second-level CP goes into a loop we could end up burning
quite a lot of CPU time and impacting your first-level users. Technically
the only way to avoid this is a suitably privileged first-level monitor,
which you probably have already if you charge your users for CPU usage.
We do not want to accept the responsibility of having a class A/B userid
on your first level system that we would have to use via TELNET over an
open network, nor do we want to get into legal battles over the extent of
our responsibility for a loop caused by a CP bug when operations and VM
maintenance/service are under your responsibility. So we will want a
mechanism whereby we can say that we refuse to be charged for more than a
certain amount of hours per month, and you will take steps to make sure
this does not happen (and, even if it does happen, we will only pay for
the amounts we registered).
We will provide free licenses for LISTSERV and LMail valid on both first
and second-level systems. This may constitute an advantage in kind if you
are otherwise paying for our products.
You are free to propose any charging schedule. We can pay for the exact
amounts used or a fixed sum for a "package" deal. We can also pay part of
the costs in kind (licenses for other machines and free access to the
mailing list service), or then give you a percentage of our profits on
the mailing list service.
|